Merge branch 'master' into compilade/mamba2

.devops/full-rocm.Dockerfile

@@ -11,7 +11,7 @@ FROM ${BASE_ROCM_DEV_CONTAINER} AS build
 # Unless otherwise specified, we make a fat build.
 # List from https://github.com/ggerganov/llama.cpp/pull/1087#issuecomment-1682807878
 # This is mostly tied to rocBLAS supported archs.
-ARG ROCM_DOCKER_ARCH=\
+ARG ROCM_DOCKER_ARCH="\
     gfx803 \
     gfx900 \
     gfx906 \
@@ -21,7 +21,7 @@ ARG ROCM_DOCKER_ARCH=\
     gfx1030 \
     gfx1100 \
     gfx1101 \
-    gfx1102
+    gfx1102"
 
 COPY requirements.txt   requirements.txt
 COPY requirements       requirements
@@ -34,7 +34,7 @@ WORKDIR /app
 COPY . .
 
 # Set nvcc architecture
-ENV GPU_TARGETS=${ROCM_DOCKER_ARCH}
+ENV AMDGPU_TARGETS=${ROCM_DOCKER_ARCH}
 # Enable ROCm
 ENV GGML_HIPBLAS=1
 ENV CC=/opt/rocm/llvm/bin/clang

.devops/llama-cli-rocm.Dockerfile

@@ -11,7 +11,7 @@ FROM ${BASE_ROCM_DEV_CONTAINER} AS build
 # Unless otherwise specified, we make a fat build.
 # List from https://github.com/ggerganov/llama.cpp/pull/1087#issuecomment-1682807878
 # This is mostly tied to rocBLAS supported archs.
-ARG ROCM_DOCKER_ARCH=\
+ARG ROCM_DOCKER_ARCH="\
     gfx803 \
     gfx900 \
     gfx906 \
@@ -21,7 +21,7 @@ ARG ROCM_DOCKER_ARCH=\
     gfx1030 \
     gfx1100 \
     gfx1101 \
-    gfx1102
+    gfx1102"
 
 COPY requirements.txt   requirements.txt
 COPY requirements       requirements
@@ -34,7 +34,7 @@ WORKDIR /app
 COPY . .
 
 # Set nvcc architecture
-ENV GPU_TARGETS=${ROCM_DOCKER_ARCH}
+ENV AMDGPU_TARGETS=${ROCM_DOCKER_ARCH}
 # Enable ROCm
 ENV GGML_HIPBLAS=1
 ENV CC=/opt/rocm/llvm/bin/clang

.devops/llama-server-rocm.Dockerfile

@@ -11,7 +11,7 @@ FROM ${BASE_ROCM_DEV_CONTAINER} AS build
 # Unless otherwise specified, we make a fat build.
 # List from https://github.com/ggerganov/llama.cpp/pull/1087#issuecomment-1682807878
 # This is mostly tied to rocBLAS supported archs.
-ARG ROCM_DOCKER_ARCH=\
+ARG ROCM_DOCKER_ARCH="\
     gfx803 \
     gfx900 \
     gfx906 \
@@ -21,7 +21,7 @@ ARG ROCM_DOCKER_ARCH=\
     gfx1030 \
     gfx1100 \
     gfx1101 \
-    gfx1102
+    gfx1102"
 
 COPY requirements.txt   requirements.txt
 COPY requirements       requirements
@@ -34,7 +34,7 @@ WORKDIR /app
 COPY . .
 
 # Set nvcc architecture
-ENV GPU_TARGETS=${ROCM_DOCKER_ARCH}
+ENV AMDGPU_TARGETS=${ROCM_DOCKER_ARCH}
 # Enable ROCm
 ENV GGML_HIPBLAS=1
 ENV CC=/opt/rocm/llvm/bin/clang

.dockerignore

@@ -1,7 +1,7 @@
 *.o
 *.a
 .cache/
-.git/
+# Do not ignore .git directory, otherwise the reported build number will always be 0
 .github/
 .gitignore
 .vs/

.github/workflows/build.yml

@@ -956,6 +956,7 @@ jobs:
           cp "${{ env.ONEAPI_ROOT }}/compiler/latest/bin/sycl7.dll" ./build/bin
           cp "${{ env.ONEAPI_ROOT }}/compiler/latest/bin/svml_dispmd.dll" ./build/bin
           cp "${{ env.ONEAPI_ROOT }}/compiler/latest/bin/libmmd.dll" ./build/bin
+          cp "${{ env.ONEAPI_ROOT }}/compiler/latest/bin/libiomp5md.dll" ./build/bin
           echo "cp oneAPI running time dll files to ./build/bin done"
           7z a llama-${{ steps.tag.outputs.name }}-bin-win-sycl-x64.zip ./build/bin/*
 
@@ -967,6 +968,7 @@ jobs:
           name: llama-bin-win-sycl-x64.zip
 
   windows-latest-cmake-hip:
+    if: ${{ github.event.inputs.create_release != 'true' }}
     runs-on: windows-latest
 
     steps:
@@ -994,8 +996,72 @@ jobs:
         run: |
           $env:HIP_PATH=$(Resolve-Path 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' | split-path | split-path)
           $env:CMAKE_PREFIX_PATH="${env:HIP_PATH}"
-          cmake -G "Unix Makefiles" -B build -S . -DCMAKE_C_COMPILER="${env:HIP_PATH}\bin\clang.exe" -DCMAKE_CXX_COMPILER="${env:HIP_PATH}\bin\clang++.exe" -DGGML_HIPBLAS=ON
+          cmake -G "Unix Makefiles" -B build -S . -DCMAKE_C_COMPILER="${env:HIP_PATH}\bin\clang.exe" -DCMAKE_CXX_COMPILER="${env:HIP_PATH}\bin\clang++.exe" -DGGML_HIPBLAS=ON -DCMAKE_BUILD_TYPE=Release -DGGML_RPC=ON
-          cmake --build build --config Release
+          cmake --build build -j ${env:NUMBER_OF_PROCESSORS}
+
+  windows-latest-cmake-hip-release:
+    if: ${{ ( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true' }}
+    runs-on: windows-latest
+
+    strategy:
+      matrix:
+        gpu_target: [gfx1100, gfx1101, gfx1030]
+
+    steps:
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v4
+
+      - name: Install
+        id: depends
+        run: |
+          $ErrorActionPreference = "Stop"
+          write-host "Downloading AMD HIP SDK Installer"
+          Invoke-WebRequest -Uri "https://download.amd.com/developer/eula/rocm-hub/AMD-Software-PRO-Edition-24.Q3-WinSvr2022-For-HIP.exe" -OutFile "${env:RUNNER_TEMP}\rocm-install.exe"
+          write-host "Installing AMD HIP SDK"
+          Start-Process "${env:RUNNER_TEMP}\rocm-install.exe" -ArgumentList '-install' -NoNewWindow -Wait
+          write-host "Completed AMD HIP SDK installation"
+
+      - name: Verify ROCm
+        id: verify
+        run: |
+          & 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' --version
+
+      - name: Build
+        id: cmake_build
+        run: |
+          $env:HIP_PATH=$(Resolve-Path 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' | split-path | split-path)
+          $env:CMAKE_PREFIX_PATH="${env:HIP_PATH}"
+          cmake -G "Unix Makefiles" -B build -S . -DCMAKE_C_COMPILER="${env:HIP_PATH}\bin\clang.exe" -DCMAKE_CXX_COMPILER="${env:HIP_PATH}\bin\clang++.exe" -DGGML_HIPBLAS=ON -DCMAKE_BUILD_TYPE=Release -DAMDGPU_TARGETS=${{ matrix.gpu_target }} -DGGML_RPC=ON
+          cmake --build build -j ${env:NUMBER_OF_PROCESSORS}
+          md "build\bin\rocblas\library\"
+          cp "${env:HIP_PATH}\bin\hipblas.dll" "build\bin\"
+          cp "${env:HIP_PATH}\bin\rocblas.dll" "build\bin\"
+          cp "${env:HIP_PATH}\bin\rocblas\library\*" "build\bin\rocblas\library\"
+
+      - name: Determine tag name
+        id: tag
+        shell: bash
+        run: |
+          BUILD_NUMBER="$(git rev-list --count HEAD)"
+          SHORT_HASH="$(git rev-parse --short=7 HEAD)"
+          if [[ "${{ env.BRANCH_NAME }}" == "master" ]]; then
+            echo "name=b${BUILD_NUMBER}" >> $GITHUB_OUTPUT
+          else
+            SAFE_NAME=$(echo "${{ env.BRANCH_NAME }}" | tr '/' '-')
+            echo "name=${SAFE_NAME}-b${BUILD_NUMBER}-${SHORT_HASH}" >> $GITHUB_OUTPUT
+          fi
+
+      - name: Pack artifacts
+        id: pack_artifacts
+        run: |
+          7z a llama-${{ steps.tag.outputs.name }}-bin-win-hip-x64-${{ matrix.gpu_target }}.zip .\build\bin\*
+
+      - name: Upload artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          path: llama-${{ steps.tag.outputs.name }}-bin-win-hip-x64-${{ matrix.gpu_target }}.zip
+          name: llama-bin-win-hip-x64-${{ matrix.gpu_target }}.zip
 
   ios-xcode-build:
     runs-on: macos-latest
@@ -1060,6 +1126,7 @@ jobs:
       - macOS-latest-cmake
       - windows-latest-cmake
       - windows-latest-cmake-cuda
+      - windows-latest-cmake-hip-release
       - macOS-latest-cmake-arm64
       - macOS-latest-cmake-x64
 

.github/workflows/docker.yml

@@ -15,11 +15,17 @@ on:
     branches:
       - master
     paths: ['.github/workflows/docker.yml', '.devops/*.Dockerfile', '**/CMakeLists.txt', '**/Makefile', '**/*.h', '**/*.hpp', '**/*.c', '**/*.cpp', '**/*.cu', '**/*.cuh', '**/*.swift', '**/*.m', '**/*.metal']
+  workflow_dispatch: # allows manual triggering, useful for debugging
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }}
   cancel-in-progress: true
 
+# Fine-grant permission
+# https://docs.github.com/en/actions/security-for-github-actions/security-guides/automatic-token-authentication#modifying-the-permissions-for-the-github_token
+permissions:
+  packages: write
+
 jobs:
   push_to_registry:
     name: Push Docker image to Docker Hub
@@ -46,6 +52,8 @@ jobs:
     steps:
       - name: Check out the repo
         uses: actions/checkout@v4
+        with:
+          fetch-depth: 0 # preserve git history, so we can determine the build number
 
       - name: Set up QEMU
         uses: docker/setup-qemu-action@v2
@@ -60,6 +68,34 @@ jobs:
           username: ${{ github.repository_owner }}
           password: ${{ secrets.GITHUB_TOKEN }}
 
+      - name: Determine tag name
+        id: tag
+        shell: bash
+        run: |
+          BUILD_NUMBER="$(git rev-list --count HEAD)"
+          SHORT_HASH="$(git rev-parse --short=7 HEAD)"
+          REPO_OWNER="${GITHUB_REPOSITORY_OWNER@L}"  # to lower case
+          REPO_NAME="${{ github.event.repository.name }}"
+
+          # determine tag name postfix (build number, commit hash)
+          if [[ "${{ env.GITHUB_BRANCH_NAME }}" == "master" ]]; then
+            TAG_POSTFIX="b${BUILD_NUMBER}"
+          else
+            SAFE_NAME=$(echo "${{ env.GITHUB_BRANCH_NAME }}" | tr '/' '-')
+            TAG_POSTFIX="${SAFE_NAME}-${SHORT_HASH}"
+          fi
+
+          # list all tags possible
+          TAGS=""
+          TAGS="${TAGS}ghcr.io/${REPO_OWNER}/${REPO_NAME}:${{ matrix.config.tag }},"
+          TAGS="${TAGS}ghcr.io/${REPO_OWNER}/${REPO_NAME}:${{ matrix.config.tag }}-${TAG_POSTFIX}"
+
+          echo "output_tags=$TAGS" >> $GITHUB_OUTPUT
+          echo "output_tags=$TAGS"  # print out for debugging
+        env:
+          GITHUB_BRANCH_NAME: ${{ github.head_ref || github.ref_name }}
+          GITHUB_REPOSITORY_OWNER: '${{ github.repository_owner }}'
+
       # https://github.com/jlumbroso/free-disk-space/tree/54081f138730dfa15788a46383842cd2f914a1be#example
       - name: Free Disk Space (Ubuntu)
         uses: jlumbroso/free-disk-space@main
@@ -77,25 +113,6 @@ jobs:
           docker-images: true
           swap-storage: true
 
-      - name: Determine tag name
-        id: tag
-        shell: bash
-        run: |
-          BUILD_NUMBER="$(git rev-list --count HEAD)"
-          SHORT_HASH="$(git rev-parse --short=7 HEAD)"
-          if [[ "${{ env.BRANCH_NAME }}" == "master" ]]; then
-            echo "name=b${BUILD_NUMBER}" >> $GITHUB_OUTPUT
-          else
-            SAFE_NAME=$(echo "${{ env.BRANCH_NAME }}" | tr '/' '-')
-            echo "name=${SAFE_NAME}-b${BUILD_NUMBER}-${SHORT_HASH}" >> $GITHUB_OUTPUT
-          fi
-
-      - name: Downcase github.repository_owner
-        run: |
-          echo "repository_owner_lowercase=${GITHUB_REPOSITORY_OWNER@L}" >> $GITHUB_ENV
-        env:
-          GITHUB_REPOSITORY_OWNER: '${{ github.repository_owner }}'
-
       - name: Build and push Docker image (tagged + versioned)
         if: github.event_name == 'push'
         uses: docker/build-push-action@v6
@@ -103,5 +120,6 @@ jobs:
           context: .
           push: true
           platforms: ${{ matrix.config.platforms }}
-          tags: "ghcr.io/${{ env.repository_owner_lowercase }}/llama.cpp:${{ matrix.config.tag }}-${{ env.COMMIT_SHA }},ghcr.io/${{ env.repository_owner_lowercase }}/llama.cpp:${{ matrix.config.tag }},ghcr.io/${{ env.repository_owner_lowercase }}/llama.cpp:${{ matrix.config.tag }}-${{ steps.tag.outputs.name }}"
+          # tag list is generated from step above
+          tags: ${{ steps.tag.outputs.output_tags }}
           file: ${{ matrix.config.dockerfile }}

.github/workflows/python-type-check.yml

@@ -4,11 +4,13 @@ on:
   push:
     paths:
       - '.github/workflows/python-type-check.yml'
+      - 'pyrightconfig.json'
       - '**.py'
       - '**/requirements*.txt'
   pull_request:
     paths:
       - '.github/workflows/python-type-check.yml'
+      - 'pyrightconfig.json'
       - '**.py'
       - '**/requirements*.txt'
 
@@ -33,6 +35,6 @@ jobs:
       - name: Type-check with Pyright
         uses: jakebailey/pyright-action@v2
         with:
-          version: 1.1.370
+          version: 1.1.382
           level: warning
           warnings: true

CMakeLists.txt

@@ -62,6 +62,9 @@ option(LLAMA_SANITIZE_THREAD    "llama: enable thread sanitizer"    OFF)
 option(LLAMA_SANITIZE_ADDRESS   "llama: enable address sanitizer"   OFF)
 option(LLAMA_SANITIZE_UNDEFINED "llama: enable undefined sanitizer" OFF)
 
+# utils
+option(LLAMA_BUILD_COMMON "llama: build common utils library" ON)
+
 # extra artifacts
 option(LLAMA_BUILD_TESTS    "llama: build tests"          ${LLAMA_STANDALONE})
 option(LLAMA_BUILD_EXAMPLES "llama: build examples"       ${LLAMA_STANDALONE})
@@ -191,15 +194,17 @@ install(FILES "${CMAKE_CURRENT_BINARY_DIR}/llama.pc"
         DESTINATION lib/pkgconfig)
 
 #
-# programs, examples and tests
+# utils, programs, examples and tests
 #
 
-add_subdirectory(common)
+if (LLAMA_BUILD_COMMON)
+    add_subdirectory(common)
+endif()
 
 if (LLAMA_BUILD_TESTS AND NOT CMAKE_JS_VERSION)
     include(CTest)
     add_subdirectory(tests)
-endif ()
+endif()
 
 if (LLAMA_BUILD_EXAMPLES)
     add_subdirectory(examples)

CONTRIBUTING.md

@@ -27,3 +27,8 @@
 
 ![matmul](media/matmul.png)
 
+# Resources
+
+The Github issues, PRs and discussions contain a lot of information that can be useful to get familiar with the codebase. For convenience, some of the more important information is referenced from Github projects:
+
+https://github.com/ggerganov/llama.cpp/projects

Makefile

@@ -611,7 +611,7 @@ ifdef GGML_CUDA
 
 		MK_CPPFLAGS  += -DGGML_USE_CUDA -I$(CUDA_PATH)/include
 		MK_LDFLAGS   += -lmusa -lmublas -lmusart -lpthread -ldl -lrt -L$(CUDA_PATH)/lib -L/usr/lib64
-		MK_NVCCFLAGS += -x musa -mtgpu --cuda-gpu-arch=mp_22
+		MK_NVCCFLAGS += -x musa -mtgpu --cuda-gpu-arch=mp_21 --cuda-gpu-arch=mp_22
 	else
 		ifneq ('', '$(wildcard /opt/cuda)')
 			CUDA_PATH ?= /opt/cuda

README.md

@@ -17,7 +17,8 @@ Inference of Meta's [LLaMA](https://arxiv.org/abs/2302.13971) model (and others)
 
 ## Hot topics
 
-- Huggingface GGUF editor: [discussion](https://github.com/ggerganov/llama.cpp/discussions/9268) | [tool](https://huggingface.co/spaces/CISCai/gguf-editor)
+- **Hugging Face Inference Endpoints now support GGUF out of the box! https://github.com/ggerganov/llama.cpp/discussions/9669**
+- Hugging Face GGUF editor: [discussion](https://github.com/ggerganov/llama.cpp/discussions/9268) | [tool](https://huggingface.co/spaces/CISCai/gguf-editor)
 
 ----
 
@@ -112,6 +113,7 @@ Typically finetunes of the base models below are supported as well.
 - Go: [go-skynet/go-llama.cpp](https://github.com/go-skynet/go-llama.cpp)
 - Node.js: [withcatai/node-llama-cpp](https://github.com/withcatai/node-llama-cpp)
 - JS/TS (llama.cpp server client): [lgrammel/modelfusion](https://modelfusion.dev/integration/model-provider/llamacpp)
+- JS/TS (Programmable Prompt Engine CLI): [offline-ai/cli](https://github.com/offline-ai/cli)
 - JavaScript/Wasm (works in browser): [tangledgroup/llama-cpp-wasm](https://github.com/tangledgroup/llama-cpp-wasm)
 - Typescript/Wasm (nicer API, available on npm): [ngxson/wllama](https://github.com/ngxson/wllama)
 - Ruby: [yoshoku/llama_cpp.rb](https://github.com/yoshoku/llama_cpp.rb)
@@ -172,6 +174,7 @@ Unless otherwise noted these projects are open-source with permissive licensing:
 **Tools:**
 
 - [akx/ggify](https://github.com/akx/ggify) – download PyTorch models from HuggingFace Hub and convert them to GGML
+- [akx/ollama-dl](https://github.com/akx/ollama-dl) – download models from the Ollama library to be used directly with llama.cpp
 - [crashr/gppm](https://github.com/crashr/gppm) – launch llama.cpp instances utilizing NVIDIA Tesla P40 or P100 GPUs with reduced idle power consumption
 - [gpustack/gguf-parser](https://github.com/gpustack/gguf-parser-go/tree/main/cmd/gguf-parser) - review/check the GGUF file and estimate the memory usage
 - [Styled Lines](https://marketplace.unity.com/packages/tools/generative-ai/styled-lines-llama-cpp-model-292902) (proprietary licensed, async wrapper of inference part for game development in Unity3d with prebuild Mobile and Web platform wrappers and a model example)
@@ -440,7 +443,7 @@ To learn more how to measure perplexity using llama.cpp, [read this documentatio
 - Contributors can open PRs
 - Collaborators can push to branches in the `llama.cpp` repo and merge PRs into the `master` branch
 - Collaborators will be invited based on contributions
-- Any help with managing issues and PRs is very appreciated!
+- Any help with managing issues, PRs and projects is very appreciated!
 - See [good first issues](https://github.com/ggerganov/llama.cpp/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22) for tasks suitable for first contributions
 - Read the [CONTRIBUTING.md](CONTRIBUTING.md) for more information
 - Make sure to read this: [Inference at the edge](https://github.com/ggerganov/llama.cpp/discussions/205)

ci/run.sh

@@ -712,6 +712,81 @@ function gg_run_embd_bge_small {
     set +e
 }
 
+function gg_sum_embd_bge_small {
+    gg_printf '### %s\n\n' "${ci}"
+
+    gg_printf 'BGE Small (BERT):\n'
+    gg_printf '- status: %s\n' "$(cat $OUT/${ci}.exit)"
+    gg_printf '- f16: \n```\n%s\n```\n' "$(cat $OUT/${ci}-tg-f16.log)"
+    gg_printf '- q8_0:\n```\n%s\n```\n' "$(cat $OUT/${ci}-tg-q8_0.log)"
+}
+
+# rerank_tiny
+
+function gg_run_rerank_tiny {
+    cd ${SRC}
+
+    gg_wget models-mnt/rerank-tiny/ https://huggingface.co/jinaai/jina-reranker-v1-tiny-en/raw/main/config.json
+    gg_wget models-mnt/rerank-tiny/ https://huggingface.co/jinaai/jina-reranker-v1-tiny-en/raw/main/tokenizer.json
+    gg_wget models-mnt/rerank-tiny/ https://huggingface.co/jinaai/jina-reranker-v1-tiny-en/raw/main/tokenizer_config.json
+    gg_wget models-mnt/rerank-tiny/ https://huggingface.co/jinaai/jina-reranker-v1-tiny-en/raw/main/special_tokens_map.json
+    gg_wget models-mnt/rerank-tiny/ https://huggingface.co/jinaai/jina-reranker-v1-tiny-en/resolve/main/pytorch_model.bin
+    gg_wget models-mnt/rerank-tiny/ https://huggingface.co/jinaai/jina-reranker-v1-tiny-en/raw/main/sentence_bert_config.json
+    gg_wget models-mnt/rerank-tiny/ https://huggingface.co/jinaai/jina-reranker-v1-tiny-en/raw/main/vocab.txt
+    gg_wget models-mnt/rerank-tiny/ https://huggingface.co/jinaai/jina-reranker-v1-tiny-en/raw/main/modules.json
+    gg_wget models-mnt/rerank-tiny/ https://huggingface.co/jinaai/jina-reranker-v1-tiny-en/raw/main/config.json
+
+    gg_wget models-mnt/rerank-tiny/1_Pooling https://huggingface.co/jinaai/jina-reranker-v1-tiny-en/raw/main/1_Pooling/config.json
+
+    path_models="../models-mnt/rerank-tiny"
+
+    rm -rf build-ci-release && mkdir build-ci-release && cd build-ci-release
+
+    set -e
+
+    (time cmake -DCMAKE_BUILD_TYPE=Release ${CMAKE_EXTRA} .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log
+    (time make -j$(nproc)                                    ) 2>&1 | tee -a $OUT/${ci}-make.log
+
+    python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf
+
+    model_f16="${path_models}/ggml-model-f16.gguf"
+
+    (time ./bin/llama-embedding --model ${model_f16}  -p "what is panda?</s><s>hi\nwhat is panda?</s><s>it's a bear\nwhat is panda?</s><s>The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China." --pooling rank --embd-normalize -1 --verbose-prompt) 2>&1 | tee -a $OUT/${ci}-rk-f16.log
+
+    # sample output
+    # rerank score 0:    0.029
+    # rerank score 1:    0.029
+    # rerank score 2:    0.135
+
+    # check that the score is in the range [$3, $4]
+    function check_score {
+        qnt="$1"
+        score=$(echo "$2" | grep -oE "[0-9]+\.[0-9]+" | tail -n 1)
+
+        if [ $(echo "$score < $3" | bc) -eq 1 ] || [ $(echo "$score > $4" | bc) -eq 1 ]; then
+            printf '  - %s @ %s (FAIL: score not in range [%s, %s])\n' "$qnt" "$score" "$3" "$4"
+            return 20
+        fi
+
+        printf '  - %s @ %s OK\n' "$qnt" "$score"
+        return 0
+    }
+
+    check_score "rerank score 0" "$(cat $OUT/${ci}-rk-f16.log | grep "rerank score 0")" "0.00" "0.05" | tee -a $OUT/${ci}-rk-f16.log
+    check_score "rerank score 1" "$(cat $OUT/${ci}-rk-f16.log | grep "rerank score 1")" "0.00" "0.05" | tee -a $OUT/${ci}-rk-f16.log
+    check_score "rerank score 2" "$(cat $OUT/${ci}-rk-f16.log | grep "rerank score 2")" "0.10" "0.15" | tee -a $OUT/${ci}-rk-f16.log
+
+    set +e
+}
+
+function gg_sum_rerank_tiny {
+    gg_printf '### %s\n\n' "${ci}"
+
+    gg_printf 'Rerank Tiny (Jina):\n'
+    gg_printf '- status: %s\n' "$(cat $OUT/${ci}.exit)"
+    gg_printf '- f16: \n```\n%s\n```\n' "$(cat $OUT/${ci}-rk-f16.log)"
+}
+
 function gg_check_build_requirements {
     if ! command -v cmake &> /dev/null; then
         gg_printf 'cmake not found, please install'
@@ -726,15 +801,6 @@ function gg_check_build_requirements {
     fi
 }
 
-function gg_sum_embd_bge_small {
-    gg_printf '### %s\n\n' "${ci}"
-
-    gg_printf 'BGE Small (BERT):\n'
-    gg_printf '- status: %s\n' "$(cat $OUT/${ci}.exit)"
-    gg_printf '- f16: \n```\n%s\n```\n' "$(cat $OUT/${ci}-tg-f16.log)"
-    gg_printf '- q8_0:\n```\n%s\n```\n' "$(cat $OUT/${ci}-tg-q8_0.log)"
-}
-
 ## main
 
 export LLAMA_LOG_PREFIX=1
@@ -762,6 +828,7 @@ test $ret -eq 0 && gg_run ctest_release
 
 if [ -z ${GG_BUILD_LOW_PERF} ]; then
     test $ret -eq 0 && gg_run embd_bge_small
+    test $ret -eq 0 && gg_run rerank_tiny
 
     if [ -z ${GG_BUILD_CLOUD} ] || [ ${GG_BUILD_EXTRA_TESTS_0} ]; then
         test $ret -eq 0 && gg_run test_scripts_debug

common/arg.cpp

@@ -284,6 +284,10 @@ static bool gpt_params_parse_ex(int argc, char ** argv, gpt_params_context & ctx
         params.kv_overrides.back().key[0] = 0;
     }
 
+    if (params.reranking && params.embedding) {
+        throw std::invalid_argument("error: either --embedding or --reranking can be specified, but not both");
+    }
+
     return true;
 }
 
@@ -391,7 +395,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
         [](gpt_params & params) {
             params.verbose_prompt = true;
         }
-    ).set_examples({LLAMA_EXAMPLE_MAIN}));
+    ));
     add_opt(llama_arg(
         {"--no-display-prompt"},
         format("don't print prompt at generation (default: %s)", !params.display_prompt ? "true" : "false"),
@@ -691,7 +695,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
         [](gpt_params & params) {
             params.ctx_shift = false;
         }
-    ).set_examples({LLAMA_EXAMPLE_MAIN}));
+    ).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_NO_CONTEXT_SHIFT"));
     add_opt(llama_arg(
         {"--chunks"}, "N",
         format("max number of chunks to process (default: %d, -1 = all)", params.n_chunks),
@@ -1093,16 +1097,17 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
         }
     ).set_sparam());
     add_opt(llama_arg(
-        {"--pooling"}, "{none,mean,cls,last}",
+        {"--pooling"}, "{none,mean,cls,last,rank}",
         "pooling type for embeddings, use model default if unspecified",
         [](gpt_params & params, const std::string & value) {
             /**/ if (value == "none") { params.pooling_type = LLAMA_POOLING_TYPE_NONE; }
             else if (value == "mean") { params.pooling_type = LLAMA_POOLING_TYPE_MEAN; }
             else if (value == "cls")  { params.pooling_type = LLAMA_POOLING_TYPE_CLS;  }
             else if (value == "last") { params.pooling_type = LLAMA_POOLING_TYPE_LAST; }
+            else if (value == "rank") { params.pooling_type = LLAMA_POOLING_TYPE_RANK; }
             else { throw std::invalid_argument("invalid value"); }
         }
-    ).set_examples({LLAMA_EXAMPLE_EMBEDDING}));
+    ).set_examples({LLAMA_EXAMPLE_EMBEDDING, LLAMA_EXAMPLE_RETRIEVAL, LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_POOLING"));
     add_opt(llama_arg(
         {"--attention"}, "{causal,non,causal}",
         "attention type for embeddings, use model default if unspecified",
@@ -1121,77 +1126,77 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
             else if (value == "yarn") { params.rope_scaling_type = LLAMA_ROPE_SCALING_TYPE_YARN; }
             else { throw std::invalid_argument("invalid value"); }
         }
-    ));
+    ).set_env("LLAMA_ARG_ROPE_SCALING_TYPE"));
     add_opt(llama_arg(
         {"--rope-scale"}, "N",
         "RoPE context scaling factor, expands context by a factor of N",
         [](gpt_params & params, const std::string & value) {
             params.rope_freq_scale = 1.0f / std::stof(value);
         }
-    ));
+    ).set_env("LLAMA_ARG_ROPE_SCALE"));
     add_opt(llama_arg(
         {"--rope-freq-base"}, "N",
         "RoPE base frequency, used by NTK-aware scaling (default: loaded from model)",
         [](gpt_params & params, const std::string & value) {
             params.rope_freq_base = std::stof(value);
         }
-    ));
+    ).set_env("LLAMA_ARG_ROPE_FREQ_BASE"));
     add_opt(llama_arg(
         {"--rope-freq-scale"}, "N",
         "RoPE frequency scaling factor, expands context by a factor of 1/N",
         [](gpt_params & params, const std::string & value) {
             params.rope_freq_scale = std::stof(value);
         }
-    ));
+    ).set_env("LLAMA_ARG_ROPE_FREQ_SCALE"));
     add_opt(llama_arg(
         {"--yarn-orig-ctx"}, "N",
         format("YaRN: original context size of model (default: %d = model training context size)", params.yarn_orig_ctx),
         [](gpt_params & params, int value) {
             params.yarn_orig_ctx = value;
         }
-    ));
+    ).set_env("LLAMA_ARG_YARN_ORIG_CTX"));
     add_opt(llama_arg(
         {"--yarn-ext-factor"}, "N",
         format("YaRN: extrapolation mix factor (default: %.1f, 0.0 = full interpolation)", (double)params.yarn_ext_factor),
         [](gpt_params & params, const std::string & value) {
             params.yarn_ext_factor = std::stof(value);
         }
-    ));
+    ).set_env("LLAMA_ARG_YARN_EXT_FACTOR"));
     add_opt(llama_arg(
         {"--yarn-attn-factor"}, "N",
         format("YaRN: scale sqrt(t) or attention magnitude (default: %.1f)", (double)params.yarn_attn_factor),
         [](gpt_params & params, const std::string & value) {
             params.yarn_attn_factor = std::stof(value);
         }
-    ));
+    ).set_env("LLAMA_ARG_YARN_ATTN_FACTOR"));
     add_opt(llama_arg(
         {"--yarn-beta-slow"}, "N",
         format("YaRN: high correction dim or alpha (default: %.1f)", (double)params.yarn_beta_slow),
         [](gpt_params & params, const std::string & value) {
             params.yarn_beta_slow = std::stof(value);
         }
-    ));
+    ).set_env("LLAMA_ARG_YARN_BETA_SLOW"));
     add_opt(llama_arg(
         {"--yarn-beta-fast"}, "N",
         format("YaRN: low correction dim or beta (default: %.1f)", (double)params.yarn_beta_fast),
         [](gpt_params & params, const std::string & value) {
             params.yarn_beta_fast = std::stof(value);
         }
-    ));
+    ).set_env("LLAMA_ARG_YARN_BETA_FAST"));
     add_opt(llama_arg(
         {"-gan", "--grp-attn-n"}, "N",
         format("group-attention factor (default: %d)", params.grp_attn_n),
         [](gpt_params & params, int value) {
             params.grp_attn_n = value;
         }
-    ));
+    ).set_env("LLAMA_ARG_GRP_ATTN_N"));
     add_opt(llama_arg(
         {"-gaw", "--grp-attn-w"}, "N",
         format("group-attention width (default: %.1f)", (double)params.grp_attn_w),
         [](gpt_params & params, int value) {
             params.grp_attn_w = value;
         }
-    ));
+    ).set_env("LLAMA_ARG_GRP_ATTN_W"));
     add_opt(llama_arg(
         {"-dkvc", "--dump-kv-cache"},
         "verbose print of the KV cache",
@@ -1205,7 +1210,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
         [](gpt_params & params) {
             params.no_kv_offload = true;
         }
-    ));
+    ).set_env("LLAMA_ARG_NO_KV_OFFLOAD"));
     add_opt(llama_arg(
         {"-ctk", "--cache-type-k"}, "TYPE",
         format("KV cache data type for K (default: %s)", params.cache_type_k.c_str()),
@@ -1213,7 +1218,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
             // TODO: get the type right here
             params.cache_type_k = value;
         }
-    ));
+    ).set_env("LLAMA_ARG_CACHE_TYPE_K"));
     add_opt(llama_arg(
         {"-ctv", "--cache-type-v"}, "TYPE",
         format("KV cache data type for V (default: %s)", params.cache_type_v.c_str()),
@@ -1221,7 +1226,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
             // TODO: get the type right here
             params.cache_type_v = value;
         }
-    ));
+    ).set_env("LLAMA_ARG_CACHE_TYPE_V"));
     add_opt(llama_arg(
         {"--perplexity", "--all-logits"},
         format("return logits for all tokens in the batch (default: %s)", params.logits_all ? "true" : "false"),
@@ -1355,7 +1360,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
         [](gpt_params & params, const std::string & value) {
             params.rpc_servers = value;
         }
-    ));
+    ).set_env("LLAMA_ARG_RPC"));
 #endif
     add_opt(llama_arg(
         {"--mlock"},
@@ -1363,14 +1368,14 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
         [](gpt_params & params) {
             params.use_mlock = true;
         }
-    ));
+    ).set_env("LLAMA_ARG_MLOCK"));
     add_opt(llama_arg(
         {"--no-mmap"},
         "do not memory-map model (slower load but may reduce pageouts if not using mlock)",
         [](gpt_params & params) {
             params.use_mmap = false;
         }
-    ));
+    ).set_env("LLAMA_ARG_NO_MMAP"));
     add_opt(llama_arg(
         {"--numa"}, "TYPE",
         "attempt optimizations that help on some NUMA systems\n"
@@ -1385,7 +1390,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
             else if (value == "numactl") { params.numa = GGML_NUMA_STRATEGY_NUMACTL; }
             else { throw std::invalid_argument("invalid value"); }
         }
-    ));
+    ).set_env("LLAMA_ARG_NUMA"));
     add_opt(llama_arg(
         {"-ngl", "--gpu-layers", "--n-gpu-layers"}, "N",
         "number of layers to store in VRAM",
@@ -1433,7 +1438,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
                 fprintf(stderr, "warning: llama.cpp was compiled without support for GPU offload. Setting the split mode has no effect.\n");
             }
         }
-    ));
+    ).set_env("LLAMA_ARG_SPLIT_MODE"));
     add_opt(llama_arg(
         {"-ts", "--tensor-split"}, "N0,N1,N2,...",
         "fraction of the model to offload to each GPU, comma-separated list of proportions, e.g. 3,1",
@@ -1460,7 +1465,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
                 fprintf(stderr, "warning: llama.cpp was compiled without support for GPU offload. Setting a tensor split has no effect.\n");
             }
         }
-    ));
+    ).set_env("LLAMA_ARG_TENSOR_SPLIT"));
     add_opt(llama_arg(
         {"-mg", "--main-gpu"}, "INDEX",
         format("the GPU to use for the model (with split-mode = none), or for intermediate results and KV (with split-mode = row) (default: %d)", params.main_gpu),
@@ -1470,7 +1475,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
                 fprintf(stderr, "warning: llama.cpp was compiled without support for GPU offload. Setting the main GPU has no effect.\n");
             }
         }
-    ));
+    ).set_env("LLAMA_ARG_MAIN_GPU"));
     add_opt(llama_arg(
         {"--check-tensors"},
         format("check model tensor data for invalid values (default: %s)", params.check_tensors ? "true" : "false"),
@@ -1533,7 +1538,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
         [](gpt_params & params, const std::string & value) {
             params.model_alias = value;
         }
-    ).set_examples({LLAMA_EXAMPLE_SERVER}));
+    ).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_ALIAS"));
     add_opt(llama_arg(
         {"-m", "--model"}, "FNAME",
         ex == LLAMA_EXAMPLE_EXPORT_LORA
@@ -1741,7 +1746,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
         [](gpt_params & params, const std::string & value) {
             params.public_path = value;
         }
-    ).set_examples({LLAMA_EXAMPLE_SERVER}));
+    ).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_STATIC_PATH"));
     add_opt(llama_arg(
         {"--embedding", "--embeddings"},
         format("restrict to only support embedding use case; use only with dedicated embedding models (default: %s)", params.embedding ? "enabled" : "disabled"),
@@ -1749,6 +1754,13 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
             params.embedding = true;
         }
     ).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_EMBEDDINGS"));
+    add_opt(llama_arg(
+        {"--reranking", "--rerank"},
+        format("enable reranking endpoint on server (default: %s)", params.reranking ? "enabled" : "disabled"),
+        [](gpt_params & params) {
+            params.reranking = true;
+        }
+    ).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_RERANKING"));
     add_opt(llama_arg(
         {"--api-key"}, "KEY",
         "API key to use for authentication (default: none)",
@@ -1779,14 +1791,14 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
         [](gpt_params & params, const std::string & value) {
             params.ssl_file_key = value;
         }
-    ).set_examples({LLAMA_EXAMPLE_SERVER}));
+    ).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_SSL_KEY_FILE"));
     add_opt(llama_arg(
         {"--ssl-cert-file"}, "FNAME",
         "path to file a PEM-encoded SSL certificate",
         [](gpt_params & params, const std::string & value) {
             params.ssl_file_cert = value;
         }
-    ).set_examples({LLAMA_EXAMPLE_SERVER}));
+    ).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_SSL_CERT_FILE"));
     add_opt(llama_arg(
         {"-to", "--timeout"}, "N",
         format("server read/write timeout in seconds (default: %d)", params.timeout_read),
@@ -1794,7 +1806,7 @@ gpt_params_context gpt_params_parser_init(gpt_params & params, llama_example ex,
             params.timeout_read  = value;
             params.timeout_write = value;
         }
-    ).set_examples({LLAMA_EXAMPLE_SERVER}));
+    ).set_examples({LLAMA_EXAMPLE_SERVER}).set_env("LLAMA_ARG_TIMEOUT"));
     add_opt(llama_arg(
         {"--threads-http"}, "N",
         format("number of threads used to process HTTP requests (default: %d)", params.n_threads_http),

common/common.cpp

@@ -1023,6 +1023,11 @@ struct llama_context_params llama_context_params_from_gpt_params(const gpt_param
     cparams.flash_attn        = params.flash_attn;
     cparams.no_perf           = params.no_perf;
 
+    if (params.reranking) {
+        cparams.embeddings    = true;
+        cparams.pooling_type  = LLAMA_POOLING_TYPE_RANK;
+    }
+
     cparams.type_k = kv_cache_type_from_str(params.cache_type_k);
     cparams.type_v = kv_cache_type_from_str(params.cache_type_v);
 
@@ -1432,6 +1437,8 @@ void llama_batch_add(
                           llama_pos   pos,
     const std::vector<llama_seq_id> & seq_ids,
                                bool   logits) {
+    GGML_ASSERT(batch.seq_id[batch.n_tokens] && "llama_batch size exceeded");
+
     batch.token   [batch.n_tokens] = id;
     batch.pos     [batch.n_tokens] = pos;
     batch.n_seq_id[batch.n_tokens] = seq_ids.size();

common/common.h

@@ -271,6 +271,7 @@ struct gpt_params {
     int32_t embd_normalize = 2;     // normalisation for embendings (-1=none, 0=max absolute int16, 1=taxicab, 2=euclidean, >2=p-norm)
     std::string embd_out   = "";    // empty = default, "array" = [[],[]...], "json" = openai style, "json+" = same "json" + cosine similarity matrix
     std::string embd_sep   = "\n";  // separator of embendings
+    bool reranking         = false; // enable reranking support on server
 
     // server params
     int32_t port           = 8080;         // server listens on this network port

common/console.cpp

@@ -94,6 +94,9 @@ namespace console {
                 simple_io = true;
             }
         }
+        if (simple_io) {
+            _setmode(_fileno(stdin), _O_U8TEXT);
+        }
 #else
         // POSIX-specific console initialization
         if (!simple_io) {

common/log.cpp

@@ -82,7 +82,7 @@ struct gpt_log_entry {
             }
         }
 
-        if (level != GGML_LOG_LEVEL_NONE && prefix) {
+        if (level != GGML_LOG_LEVEL_NONE && level != GGML_LOG_LEVEL_CONT && prefix) {
             if (timestamp) {
                 // [M.s.ms.us]
                 fprintf(fcur, "%s%d.%02d.%03d.%03d%s ",

common/log.h

@@ -83,8 +83,10 @@ void gpt_log_set_timestamps(struct gpt_log * log,       bool   timestamps); // w
 #define LOG_WRN(...) LOG_TMPL(GGML_LOG_LEVEL_WARN,  0,                 __VA_ARGS__)
 #define LOG_ERR(...) LOG_TMPL(GGML_LOG_LEVEL_ERROR, 0,                 __VA_ARGS__)
 #define LOG_DBG(...) LOG_TMPL(GGML_LOG_LEVEL_DEBUG, LOG_DEFAULT_DEBUG, __VA_ARGS__)
+#define LOG_CNT(...) LOG_TMPL(GGML_LOG_LEVEL_CONT,  0,                 __VA_ARGS__)
 
 #define LOG_INFV(verbosity, ...) LOG_TMPL(GGML_LOG_LEVEL_INFO,  verbosity, __VA_ARGS__)
 #define LOG_WRNV(verbosity, ...) LOG_TMPL(GGML_LOG_LEVEL_WARN,  verbosity, __VA_ARGS__)
 #define LOG_ERRV(verbosity, ...) LOG_TMPL(GGML_LOG_LEVEL_ERROR, verbosity, __VA_ARGS__)
 #define LOG_DBGV(verbosity, ...) LOG_TMPL(GGML_LOG_LEVEL_DEBUG, verbosity, __VA_ARGS__)
+#define LOG_CNTV(verbosity, ...) LOG_TMPL(GGML_LOG_LEVEL_CONT,  verbosity, __VA_ARGS__)

common/sampling.cpp

@@ -209,7 +209,15 @@ struct gpt_sampler * gpt_sampler_init(const struct llama_model * model, const st
             GGML_ASSERT(false && "unknown mirostat version");
         }
     } else {
+        if (params.n_probs > 0) {
+            // some use cases require to sample greedily, but still obtain the probabilities of the top tokens
+            // ref: https://github.com/ggerganov/llama.cpp/pull/9605
+            //
+            // the following will not produce exactly the same probs as applyging softmax to the full vocabulary, but
+            // it is much faster, since we avoid sorting all tokens and should give a good approximation
+            llama_sampler_chain_add(result->chain, llama_sampler_init_top_k(params.n_probs));
             llama_sampler_chain_add(result->chain, llama_sampler_init_softmax());
+        }
         llama_sampler_chain_add(result->chain, llama_sampler_init_greedy());
     }
 

convert_hf_to_gguf.py

@@ -15,6 +15,7 @@ from enum import IntEnum
 from pathlib import Path
 from hashlib import sha256
 from typing import TYPE_CHECKING, Any, Callable, ContextManager, Iterable, Iterator, Literal, Sequence, TypeVar, cast
+from itertools import chain
 
 import math
 import numpy as np
@@ -64,7 +65,6 @@ class Model:
     model_name: str | None
     metadata_override: Path | None
     dir_model_card: Path
-    is_lora: bool
 
     # subclasses should define this!
     model_arch: gguf.MODEL_ARCH
@@ -72,7 +72,7 @@ class Model:
     def __init__(self, dir_model: Path, ftype: gguf.LlamaFileType, fname_out: Path, is_big_endian: bool = False,
                  use_temp_file: bool = False, eager: bool = False,
                  metadata_override: Path | None = None, model_name: str | None = None,
-                 split_max_tensors: int = 0, split_max_size: int = 0, dry_run: bool = False, small_first_shard: bool = False, is_lora: bool = False):
+                 split_max_tensors: int = 0, split_max_size: int = 0, dry_run: bool = False, small_first_shard: bool = False):
         if type(self) is Model:
             raise TypeError(f"{type(self).__name__!r} should not be directly instantiated")
 
@@ -94,7 +94,6 @@ class Model:
         self.metadata_override = metadata_override
         self.model_name = model_name
         self.dir_model_card = dir_model  # overridden in convert_lora_to_gguf.py
-        self.is_lora = is_lora  # true if model is used inside convert_lora_to_gguf.py
 
         # Apply heuristics to figure out typical tensor encoding based on first layer tensor encoding type
         if self.ftype == gguf.LlamaFileType.GUESSED:
@@ -270,10 +269,14 @@ class Model:
 
         return False
 
+    # some models need extra generated tensors (like rope_freqs)
+    def generate_extra_tensors(self) -> Iterable[tuple[str, Tensor]]:
+        return ()
+
     def prepare_tensors(self):
         max_name_len = max(len(s) for _, s in self.tensor_map.mapping.values()) + len(".weight,")
 
-        for name, data_torch in self.get_tensors():
+        for name, data_torch in chain(self.generate_extra_tensors(), self.get_tensors()):
             # we don't need these
             if name.endswith((".attention.masked_bias", ".attention.bias", ".rotary_emb.inv_freq")):
                 continue
@@ -291,8 +294,13 @@ class Model:
                     bid = int(part)
                     break
 
-            for new_name, data in ((n, d.squeeze().numpy()) for n, d in self.modify_tensors(data_torch, name, bid)):
+            for new_name, data_torch in (self.modify_tensors(data_torch, name, bid)):
-                data: np.ndarray  # type hint
+                data = data_torch.squeeze().numpy()
+
+                # if data ends up empty, it means data_torch was a scalar tensor -> restore
+                if len(data.shape) == 0:
+                    data = data_torch.numpy()
+
                 n_dims = len(data.shape)
                 data_qtype: gguf.GGMLQuantizationType | bool = self.tensor_force_quant(name, new_name, bid, n_dims)
 
@@ -592,6 +600,9 @@ class Model:
         if chkhsh == "a8594e3edff7c29c003940395316294b2c623e09894deebbc65f33f1515df79e":
             # ref: https://huggingface.co/databricks/dbrx-base
             res = "dbrx"
+        if chkhsh == "c7699093ba4255a91e702aa38a596aa81669f3525dae06c2953267dde580f448":
+            # ref: https://huggingface.co/jinaai/jina-reranker-v1-tiny-en
+            res = "jina-v1-en"
         if chkhsh == "0876d13b50744004aa9aeae05e7b0647eac9d801b5ba4668afc01e709c15e19f":
             # ref: https://huggingface.co/jinaai/jina-embeddings-v2-base-en
             res = "jina-v2-en"
@@ -640,6 +651,9 @@ class Model:
         if chkhsh == "fcace8b9cac38ce847670c970cd5892031a753a1ef381abd1d9af00f713da085":
             # ref: https://huggingface.co/microsoft/phi-2
             res = "phi-2"
+        if chkhsh == "60824e3c0d9401f89943cbb2fff727f0e2d4c545ba4df2d6e4f09a6db0f5b450":
+            # ref: https://huggingface.co/facebook/chameleon-7b
+            res = "chameleon"
 
         if res is None:
             logger.warning("\n")
@@ -1606,7 +1620,7 @@ class LlamaModel(Model):
 
         return [(self.map_tensor_name(name), data_torch)]
 
-    def prepare_tensors(self):
+    def generate_extra_tensors(self) -> Iterable[tuple[str, Tensor]]:
         if rope_scaling := self.find_hparam(["rope_scaling"], optional=True):
             if rope_scaling.get("rope_type", '').lower() == "llama3":
                 base = self.hparams.get("rope_theta", 10000.0)
@@ -1633,9 +1647,9 @@ class LlamaModel(Model):
                         smooth = (old_context_len / wavelen - low_freq_factor) / (high_freq_factor - low_freq_factor)
                         rope_factors.append(1 / ((1 - smooth) / factor + smooth))
 
-                if not self.is_lora:
+                yield (self.format_tensor_name(gguf.MODEL_TENSOR.ROPE_FREQS), torch.tensor(rope_factors, dtype=torch.float32))
-                    self.gguf_writer.add_tensor(self.format_tensor_name(gguf.MODEL_TENSOR.ROPE_FREQS), np.array(rope_factors, dtype=np.float32))
 
+    def prepare_tensors(self):
         super().prepare_tensors()
 
         if self._experts is not None:
@@ -1859,8 +1873,6 @@ class MiniCPM3Model(Model):
     def set_gguf_parameters(self):
         hparams = self.hparams
 
-        rope_dims = hparams["qk_rope_head_dim"]
-
         self.gguf_writer.add_file_type(self.ftype)
         self.gguf_writer.add_context_length(hparams["max_position_embeddings"])
         self.gguf_writer.add_embedding_length(hparams["hidden_size"])
@@ -1876,9 +1888,10 @@ class MiniCPM3Model(Model):
         self.gguf_writer.add_key_length(hparams["qk_nope_head_dim"] + hparams["qk_rope_head_dim"])
         self.gguf_writer.add_rope_dimension_count(hparams["qk_rope_head_dim"])
 
+    def generate_extra_tensors(self) -> Iterable[tuple[str, Tensor]]:
         rope_scaling = self.find_hparam(['rope_scaling'], True)
-        if rope_scaling is None:
+        if rope_scaling is not None:
-            return
+            rope_dims = self.hparams["qk_rope_head_dim"]
 
             long_factors = rope_scaling.get('long_factor', None)
             short_factors = rope_scaling.get('short_factor', None)
@@ -1889,11 +1902,11 @@ class MiniCPM3Model(Model):
             if len(long_factors) != len(short_factors) or len(long_factors) != rope_dims / 2:
                 raise ValueError(f'The length of rope long and short factors must be {rope_dims / 2}')
 
-        self.gguf_writer.add_tensor(gguf.TENSOR_NAMES[gguf.MODEL_TENSOR.ROPE_FACTORS_LONG]  + ".weight", np.array(long_factors, dtype=np.float32))
+            yield (self.format_tensor_name(gguf.MODEL_TENSOR.ROPE_FACTORS_LONG), torch.tensor(long_factors, dtype=torch.float32))
-        self.gguf_writer.add_tensor(gguf.TENSOR_NAMES[gguf.MODEL_TENSOR.ROPE_FACTORS_SHORT] + ".weight", np.array(short_factors, dtype=np.float32))
+            yield (self.format_tensor_name(gguf.MODEL_TENSOR.ROPE_FACTORS_SHORT), torch.tensor(short_factors, dtype=torch.float32))
 
     def set_vocab(self):
-        self._set_vocab_llama_hf()
+        self._set_vocab_sentencepiece()
 
     def _reverse_hf_permute(self, weights: Tensor, n_head: int, n_kv_head: int | None = None) -> Tensor:
         if n_kv_head is not None and n_head != n_kv_head:
@@ -2205,6 +2218,13 @@ class Phi3MiniModel(Model):
         self.gguf_writer.add_file_type(self.ftype)
         self.gguf_writer.add_sliding_window(self.find_hparam(["sliding_window"]))
 
+    def generate_extra_tensors(self) -> Iterable[tuple[str, Tensor]]:
+        n_embd = self.find_hparam(["hidden_size", "n_embd"])
+        n_head = self.find_hparam(["num_attention_heads", "n_head"])
+        max_pos_embds = self.find_hparam(["n_positions", "max_position_embeddings"])
+        orig_max_pos_embds = self.find_hparam(["original_max_position_embeddings"])
+        rope_dims = n_embd // n_head
+
         # write rope scaling for long context (128k) model
         rope_scaling = self.find_hparam(['rope_scaling'], True)
         if rope_scaling is None:
@@ -2234,9 +2254,8 @@ class Phi3MiniModel(Model):
         if len(long_factors) != len(short_factors) or len(long_factors) != rope_dims / 2:
             raise ValueError(f'The length of rope long and short factors must be {rope_dims / 2}')
 
-        if not self.is_lora:
+        yield (self.format_tensor_name(gguf.MODEL_TENSOR.ROPE_FACTORS_LONG), torch.tensor(long_factors, dtype=torch.float32))
-            self.gguf_writer.add_tensor(gguf.TENSOR_NAMES[gguf.MODEL_TENSOR.ROPE_FACTORS_LONG]  + ".weight", np.array(long_factors, dtype=np.float32))
+        yield (self.format_tensor_name(gguf.MODEL_TENSOR.ROPE_FACTORS_SHORT), torch.tensor(short_factors, dtype=torch.float32))
-            self.gguf_writer.add_tensor(gguf.TENSOR_NAMES[gguf.MODEL_TENSOR.ROPE_FACTORS_SHORT] + ".weight", np.array(short_factors, dtype=np.float32))
 
 
 @Model.register("PlamoForCausalLM")
@@ -2598,7 +2617,7 @@ class NomicBertModel(BertModel):
         self.gguf_writer.add_rope_freq_base(self.hparams["rotary_emb_base"])
 
 
-@Model.register("XLMRobertaModel")
+@Model.register("XLMRobertaModel", "XLMRobertaForSequenceClassification")
 class XLMRobertaModel(BertModel):
     model_arch = gguf.MODEL_ARCH.BERT
 
@@ -2696,6 +2715,11 @@ class XLMRobertaModel(BertModel):
         self.gguf_writer.add_add_eos_token(True)
 
     def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        # if name starts with "roberta.", remove the prefix
+        # e.g. https://huggingface.co/BAAI/bge-reranker-v2-m3/tree/main
+        if name.startswith("roberta."):
+            name = name[8:]
+
         # position embeddings start at pad_token_id + 1, so just chop down the weight tensor
         if name == "embeddings.position_embeddings.weight":
             if self._position_offset is not None:
@@ -3178,6 +3202,14 @@ class JinaBertV2Model(BertModel):
         self.gguf_writer.add_add_bos_token(True)
         self.gguf_writer.add_add_eos_token(True)
 
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        # if name starts with "bert.", remove the prefix
+        # e.g. https://huggingface.co/jinaai/jina-reranker-v1-tiny-en
+        if name.startswith("bert."):
+            name = name[5:]
+
+        return super().modify_tensors(data_torch, name, bid)
+
 
 @Model.register("OpenELMForCausalLM")
 class OpenELMModel(Model):
@@ -4118,7 +4150,7 @@ class ExaoneModel(Model):
                 self.gguf_writer.add_rope_scaling_type(gguf.RopeScalingType.LINEAR)
                 self.gguf_writer.add_rope_scaling_factor(hparams["rope_scaling"]["factor"])
 
-    def prepare_tensors(self):
+    def generate_extra_tensors(self) -> Iterable[tuple[str, Tensor]]:
         if rope_scaling := self.find_hparam(["rope_scaling"], optional=True):
             if rope_scaling.get("rope_type", '').lower() == "llama3":
                 base = self.hparams.get("rope_theta", 10000.0)
@@ -4145,10 +4177,7 @@ class ExaoneModel(Model):
                         smooth = (old_context_len / wavelen - low_freq_factor) / (high_freq_factor - low_freq_factor)
                         rope_factors.append(1 / ((1 - smooth) / factor + smooth))
 
-                if not self.is_lora:
+                yield (self.format_tensor_name(gguf.MODEL_TENSOR.ROPE_FREQS), torch.tensor(rope_factors, dtype=torch.float32))
-                    self.gguf_writer.add_tensor(self.format_tensor_name(gguf.MODEL_TENSOR.ROPE_FREQS), np.array(rope_factors, dtype=np.float32))
-
-        super().prepare_tensors()
 
 
 @Model.register("GraniteForCausalLM")
@@ -4173,16 +4202,87 @@ class GraniteModel(LlamaModel):
         #   consistency
         if attention_scale := self.hparams.get("attention_multiplier"):
             self.gguf_writer.add_attention_scale(attention_scale)
+            logger.info("gguf: (granite) attention_scale = %s", attention_scale)
         if embedding_scale := self.hparams.get("embedding_multiplier"):
             self.gguf_writer.add_embedding_scale(embedding_scale)
+            logger.info("gguf: (granite) embedding_scale = %s", embedding_scale)
         if residual_scale := self.hparams.get("residual_multiplier"):
             self.gguf_writer.add_residual_scale(residual_scale)
-        if logits_scaling := self.hparams.get("logits_scaling"):
+            logger.info("gguf: (granite) residual_scale = %s", residual_scale)
-            self.gguf_writer.add_logit_scale(logits_scaling)
+        if logits_scale := self.hparams.get("logits_scaling"):
+            self.gguf_writer.add_logit_scale(logits_scale)
+            logger.info("gguf: (granite) logits_scale = %s", logits_scale)
+
+
+@Model.register("GraniteMoeForCausalLM")
+class GraniteMoeModel(GraniteModel):
+    """Conversion for IBM's GraniteMoeForCausalLM"""
+    model_arch = gguf.MODEL_ARCH.GRANITE_MOE
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        """In modeling_granitemoe, the JetMoe implementation of parallel experts
+        is used. This essentially merges w1 and w3 into a single tensor with 2x
+        the hidden size that is then split during forward. To keep compatibility
+        with existing mixtral support, we pull them apart here.
+        """
+
+        if name.endswith("block_sparse_moe.input_linear.weight"):
+            ffn_dim = self.hparams["intermediate_size"]
+            assert data_torch.shape[-2] == 2 * ffn_dim, "Merged FFN tensor size must be 2 * intermediate_size"
+            gate, up = data_torch[..., :ffn_dim, :], data_torch[..., ffn_dim:, :]
+            return [
+                (self.format_tensor_name(gguf.MODEL_TENSOR.FFN_GATE_EXP, bid), gate),
+                (self.format_tensor_name(gguf.MODEL_TENSOR.FFN_UP_EXP, bid), up),
+            ]
+
+        return super().modify_tensors(data_torch, name, bid)
+
+
+@Model.register("ChameleonForConditionalGeneration")
+@Model.register("ChameleonForCausalLM")  # obsolete
+class ChameleonModel(Model):
+    model_arch = gguf.MODEL_ARCH.CHAMELEON
+
+    def set_gguf_parameters(self):
+        super().set_gguf_parameters()
+        self.gguf_writer.add_swin_norm(self.hparams.get("swin_norm", False))
+
+    def set_vocab(self):
+        self._set_vocab_gpt2()
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        # ignore image tokenizer for now
+        # TODO: remove this once image support is implemented for Chameleon
+        if name.startswith("model.vqmodel"):
+            return []
+
+        n_head = self.hparams["num_attention_heads"]
+        n_kv_head = self.hparams.get("num_key_value_heads")
+        hidden_dim = self.hparams.get("hidden_size")
+
+        if name.endswith(("q_proj.weight", "q_proj.bias")):
+            data_torch = LlamaModel.permute(data_torch, n_head, n_head)
+        if name.endswith(("k_proj.weight", "k_proj.bias")):
+            data_torch = LlamaModel.permute(data_torch, n_head, n_kv_head)
+        if name.endswith(("q_norm.weight", "q_norm.bias")):
+            data_torch = ChameleonModel._reverse_hf_permute(data_torch, n_head, hidden_dim)
+        if name.endswith(("k_norm.weight", "k_norm.bias")):
+            data_torch = ChameleonModel._reverse_hf_permute(data_torch, n_kv_head, hidden_dim)
+
+        return [(self.map_tensor_name(name), data_torch)]
+
+    # see: https://github.com/huggingface/transformers/blob/72fb02c47dbbe1999ae105319f24631cad6e2e00/src/transformers/models/chameleon/convert_chameleon_weights_to_hf.py#L176-L203
+    @staticmethod
+    def _reverse_hf_permute(data_torch, n_heads, hidden_dim):
+        head_dim = hidden_dim // n_heads
+        data_torch = data_torch[0].view(2, head_dim // 2).t().reshape(1, -1)
+        data_torch = data_torch.repeat_interleave(n_heads, 0)
+        return data_torch
 
 
 ###### CONVERSION LOGIC ######
 
+
 # tree of lazy tensors
 class LazyTorchTensor(gguf.LazyBase):
     _tensor_type = torch.Tensor

convert_hf_to_gguf_update.py

@@ -81,6 +81,7 @@ models = [
     {"name": "qwen2",          "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/Qwen/Qwen1.5-7B", },
     {"name": "olmo",           "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/allenai/OLMo-1.7-7B-hf", },
     {"name": "dbrx",           "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/databricks/dbrx-base", },
+    {"name": "jina-v1-en",     "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-reranker-v1-tiny-en", },
     {"name": "jina-v2-en",     "tokt": TOKENIZER_TYPE.WPM, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-en", }, # WPM!
     {"name": "jina-v2-es",     "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-es", },
     {"name": "jina-v2-de",     "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-de", },
@@ -99,6 +100,7 @@ models = [
     {'name': "gpt3-finnish",   "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/TurkuNLP/gpt3-finnish-small", },
     {"name": "exaone",         "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct", },
     {"name": "phi-2",          "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/microsoft/phi-2", },
+    {"name": "chameleon",      "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/facebook/chameleon-7b", },
 ]
 
 

convert_lora_to_gguf.py

@@ -331,6 +331,10 @@ if __name__ == '__main__':
                 self.gguf_writer.add_float32(gguf.Keys.Adapter.LORA_ALPHA, self.lora_alpha)
                 super().set_gguf_parameters()
 
+            def generate_extra_tensors(self) -> Iterable[tuple[str, Tensor]]:
+                # Never add extra tensors (e.g. rope_freqs) for LoRA adapters
+                return ()
+
             def get_tensors(self) -> Iterator[tuple[str, Tensor]]:
                 tensor_map: dict[str, PartialLoraTensor] = {}
 
@@ -392,7 +396,6 @@ if __name__ == '__main__':
             dry_run=args.dry_run,
             dir_lora_model=dir_lora,
             lora_alpha=alpha,
-            is_lora=True,
         )
 
         logger.info("Exporting model...")

examples/convert-llama2c-to-ggml/convert-llama2c-to-ggml.cpp

@@ -201,7 +201,7 @@ static void print_sample_weights(TransformerWeights *w){
 
 //////////////////////////////////////// ggml structs and functions required to load models, configs and save the model.
 
-struct llama_vocab {
+struct my_llama_vocab {
     using id    = int32_t;
     using token = std::string;
     using ttype = llama_token_type;
@@ -525,7 +525,7 @@ static std::string llama_escape_whitespaces(const std::string & text) {
     return out.str();
 }
 
-static void load_vocab(const char * filename, const Config * config, struct llama_vocab * vocab) {
+static void load_vocab(const char * filename, const Config * config, struct my_llama_vocab * vocab) {
     if (is_ggml_file(filename)) {
         LOG_INF("%s: Loading vocabulary from gguf file %s\n", __func__, filename);
         struct ggml_context * ctx_data = NULL;
@@ -583,13 +583,13 @@ static void load_vocab(const char * filename, const Config * config, struct llam
         const int  n_vocab = config->vocab_size;
         /* uint32_t max_token_length =  */ file.read_u32(); // unused
         vocab->id_to_token.resize(n_vocab);
-        for (llama_vocab::id id=0; id<n_vocab; ++id) {
+        for (my_llama_vocab::id id=0; id<n_vocab; ++id) {
             float_t score = file.read_f32();
             uint32_t len = file.read_u32();
             std::string text = file.read_string(len);
 
             unsigned char byte_val;
-            llama_vocab::ttype type = LLAMA_TOKEN_TYPE_NORMAL;
+            my_llama_vocab::ttype type = LLAMA_TOKEN_TYPE_NORMAL;
             if (id == UNKNOWN_TOKEN_ID) {
                 text = "<unk>";
                 type = LLAMA_TOKEN_TYPE_UNKNOWN;
@@ -631,7 +631,7 @@ static void convert_weights_ak_to_gg(struct ggml_tensor * gg_weights, const floa
 }
 
 static void save_as_llama_model(
-    struct llama_vocab * vocab, struct my_llama_model * model, TransformerWeights* w, const char * filename
+    struct my_llama_vocab * vocab, struct my_llama_model * model, TransformerWeights* w, const char * filename
 ) {
     // convert AK weights into GG weights one by one.
     // w->token_embedding_table -> model->tok_embeddings
@@ -671,7 +671,7 @@ static void save_as_llama_model(
     std::vector<const char*> tokens;
     std::vector<float> scores;
     std::vector<llama_token_type> token_types;
-    for (const llama_vocab::token_data & token_data : vocab->id_to_token) {
+    for (const my_llama_vocab::token_data & token_data : vocab->id_to_token) {
         tokens.push_back(token_data.text.c_str());
         scores.push_back(token_data.score);
         token_types.push_back(token_data.type);
@@ -905,7 +905,7 @@ int main(int argc, char ** argv) {
         fclose(file);
     }
 
-    struct llama_vocab vocab;
+    struct my_llama_vocab vocab;
     load_vocab(params.fn_vocab_model, &config, &vocab);
 
     struct my_llama_model model;

examples/cvector-generator/pca.hpp

@@ -204,13 +204,6 @@ static ggml_status compute_piter(
         ggml_backend_cpu_set_n_threads(model.backend, params.n_threads);
     }
 
-// TODO: enable GPU support when support for GGML_OP_SQRT is added
-//#ifdef GGML_USE_METAL
-//    if (ggml_backend_is_metal(model.backend)) {
-//        ggml_backend_metal_set_n_cb(model.backend, params.n_threads);
-//    }
-//#endif
-
     ggml_status res = ggml_backend_graph_compute(model.backend, gf);
     if (res == GGML_STATUS_SUCCESS) {
         auto extract_i = [](std::string prefix, std::string str) -> int {

examples/embedding/embedding.cpp

@@ -135,7 +135,7 @@ int main(int argc, char ** argv) {
     // tokenize the prompts and trim
     std::vector<std::vector<int32_t>> inputs;
     for (const auto & prompt : prompts) {
-        auto inp = ::llama_tokenize(ctx, prompt, true, false);
+        auto inp = ::llama_tokenize(ctx, prompt, true, true);
         if (inp.size() > n_batch) {
             LOG_ERR("%s: number of tokens in input line (%lld) exceeds batch size (%lld), increase batch size and re-run\n",
                     __func__, (long long int) inp.size(), (long long int) n_batch);
@@ -234,6 +234,11 @@ int main(int argc, char ** argv) {
                 }
                 LOG("\n");
             }
+        } else if (pooling_type == LLAMA_POOLING_TYPE_RANK) {
+            for (int j = 0; j < n_embd_count; j++) {
+                // NOTE: if you change this log - update the tests in ci/run.sh
+                LOG("rerank score %d: %8.3f\n", j, emb[j * n_embd]);
+            }
         } else {
             // print the first part of the embeddings or for a single prompt, the full embedding
             for (int j = 0; j < n_prompts; j++) {

examples/gen-docs/gen-docs.cpp

@@ -6,15 +6,12 @@
 
 // Export usage message (-h) to markdown format
 
-static void export_md(std::string fname, llama_example ex) {
+static void write_table_header(std::ofstream & file) {
-    std::ofstream file(fname, std::ofstream::out | std::ofstream::trunc);
-
-    gpt_params params;
-    auto ctx_arg = gpt_params_parser_init(params, ex);
-
     file << "| Argument | Explanation |\n";
     file << "| -------- | ----------- |\n";
-    for (auto & opt : ctx_arg.options) {
+}
+
+static void write_table_entry(std::ofstream & file, const llama_arg & opt) {
     file << "| `";
     // args
     for (const auto & arg : opt.args) {
@@ -41,9 +38,43 @@ static void export_md(std::string fname, llama_example ex) {
     string_replace_all(md_help, "\n", "<br/>");
     string_replace_all(md_help, "|", "\\|");
     file << "` | " << md_help << " |\n";
+}
+
+static void write_table(std::ofstream & file, std::vector<llama_arg *> & opts) {
+    write_table_header(file);
+    for (const auto & opt : opts) {
+        write_table_entry(file, *opt);
     }
 }
 
+static void export_md(std::string fname, llama_example ex) {
+    std::ofstream file(fname, std::ofstream::out | std::ofstream::trunc);
+
+    gpt_params params;
+    auto ctx_arg = gpt_params_parser_init(params, ex);
+
+    std::vector<llama_arg *> common_options;
+    std::vector<llama_arg *> sparam_options;
+    std::vector<llama_arg *> specific_options;
+    for (auto & opt : ctx_arg.options) {
+        // in case multiple LLAMA_EXAMPLE_* are set, we prioritize the LLAMA_EXAMPLE_* matching current example
+        if (opt.is_sparam) {
+            sparam_options.push_back(&opt);
+        } else if (opt.in_example(ctx_arg.ex)) {
+            specific_options.push_back(&opt);
+        } else {
+            common_options.push_back(&opt);
+        }
+    }
+
+    file << "**Common params**\n\n";
+    write_table(file, common_options);
+    file << "\n\n**Sampling params**\n\n";
+    write_table(file, sparam_options);
+    file << "\n\n**Example-specific params**\n\n";
+    write_table(file, specific_options);
+}
+
 int main(int, char **) {
     export_md("autogen-main.md", LLAMA_EXAMPLE_MAIN);
     export_md("autogen-server.md", LLAMA_EXAMPLE_SERVER);

examples/imatrix/imatrix.cpp

@@ -572,6 +572,7 @@ int main(int argc, char ** argv) {
 
     params.n_ctx = 512;
     params.logits_all = true;
+    params.escape = false;
 
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_IMATRIX, print_usage)) {
         return 1;

examples/infill/infill.cpp

@@ -97,6 +97,11 @@ static void sigint_handler(int signo) {
             LOG("\n");
             gpt_perf_print(*g_ctx, *g_smpl);
             write_logfile(*g_ctx, *g_params, *g_model, *g_input_tokens, g_output_ss->str(), *g_output_tokens);
+
+            // make sure all logs are flushed
+            LOG("Interrupted by user\n");
+            gpt_log_pause(gpt_log_main());
+
             _exit(130);
         }
     }
@@ -258,9 +263,9 @@ int main(int argc, char ** argv) {
         if (params.n_keep > 0) {
         LOG_INF("%s: static prompt based on n_keep: '", __func__);
             for (int i = 0; i < params.n_keep; i++) {
-                LOG("%s", llama_token_to_piece(ctx, embd_inp[i]).c_str());
+                LOG_CNT("%s", llama_token_to_piece(ctx, embd_inp[i]).c_str());
             }
-            LOG("'\n");
+            LOG_CNT("'\n");
         }
         LOG_INF("\n");
     }
@@ -301,8 +306,8 @@ int main(int argc, char ** argv) {
 
     LOG_INF("generate: n_ctx = %d, n_batch = %d, n_predict = %d, n_keep = %d\n", n_ctx, params.n_batch, params.n_predict, params.n_keep);
 
-    LOG("\n");
+    LOG_INF("\n");
-    LOG("\n#####  Infill mode  #####\n\n");
+    LOG_INF("\n#####  Infill mode  #####\n\n");
     if (params.interactive) {
         const char *control_message;
         if (params.multiline_input) {
@@ -313,11 +318,11 @@ int main(int argc, char ** argv) {
                               " - To return control without starting a new line, end your input with '/'.\n"
                               " - If you want to submit another line, end your input with '\\'.\n";
         }
-        LOG("== Running in interactive mode. ==\n");
+        LOG_INF("== Running in interactive mode. ==\n");
 #if defined (__unix__) || (defined (__APPLE__) && defined (__MACH__)) || defined (_WIN32)
-        LOG(       " - Press Ctrl+C to interject at any time.\n");
+        LOG_INF(       " - Press Ctrl+C to interject at any time.\n");
 #endif
-        LOG(       "%s\n", control_message);
+        LOG_INF(       "%s\n", control_message);
 
         is_interacting = params.interactive_first;
     }

examples/llava/clip.cpp

@@ -2444,12 +2444,6 @@ bool clip_image_batch_encode(clip_ctx * ctx, const int n_threads, const clip_ima
         ggml_backend_cpu_set_n_threads(ctx->backend, n_threads);
     }
 
-#ifdef GGML_USE_METAL
-    if (ggml_backend_is_metal(ctx->backend)) {
-        ggml_backend_metal_set_n_cb(ctx->backend, n_threads);
-    }
-#endif
-
     ggml_backend_graph_compute(ctx->backend, gf);
 
     // the last node is the embedding tensor

examples/llava/convert_image_encoder_to_gguf.py

@@ -274,7 +274,7 @@ fout.add_bool("clip.use_gelu", use_gelu)
 
 
 if has_llava_projector:
-    model.vision_model.encoder.layers.pop(-1)  # pyright: ignore[reportAttributeAccessIssue]
+    model.vision_model.encoder.layers.pop(-1)
     projector = torch.load(args.llava_projector)
     for name, data in projector.items():
         name = get_tensor_name(name)
@@ -288,7 +288,7 @@ if has_llava_projector:
 
     print("Projector tensors added\n")
 
-state_dict = model.state_dict()  # pyright: ignore[reportAttributeAccessIssue]
+state_dict = model.state_dict()
 for name, data in state_dict.items():
     if should_skip_tensor(name, has_text_encoder, has_vision_encoder, has_llava_projector):
         # we don't need this

examples/main/main.cpp

@@ -116,6 +116,11 @@ static void sigint_handler(int signo) {
             LOG("\n");
             gpt_perf_print(*g_ctx, *g_smpl);
             write_logfile(*g_ctx, *g_params, *g_model, *g_input_tokens, g_output_ss->str(), *g_output_tokens);
+
+            // make sure all logs are flushed
+            LOG("Interrupted by user\n");
+            gpt_log_pause(gpt_log_main());
+
             _exit(130);
         }
     }
@@ -380,9 +385,9 @@ int main(int argc, char ** argv) {
         if (params.n_keep > add_bos) {
             LOG_INF("%s: static prompt based on n_keep: '", __func__);
             for (int i = 0; i < params.n_keep; i++) {
-                LOG("%s", llama_token_to_piece(ctx, embd_inp[i]).c_str());
+                LOG_CNT("%s", llama_token_to_piece(ctx, embd_inp[i]).c_str());
             }
-            LOG("'\n");
+            LOG_CNT("'\n");
         }
         LOG_INF("\n");
     }
@@ -404,40 +409,40 @@ int main(int argc, char ** argv) {
     }
 
     if (params.interactive) {
-        LOG("%s: interactive mode on.\n", __func__);
+        LOG_INF("%s: interactive mode on.\n", __func__);
 
         if (!params.antiprompt.empty()) {
             for (const auto & antiprompt : params.antiprompt) {
-                LOG("Reverse prompt: '%s'\n", antiprompt.c_str());
+                LOG_INF("Reverse prompt: '%s'\n", antiprompt.c_str());
                 if (params.verbose_prompt) {
                     auto tmp = ::llama_tokenize(ctx, antiprompt, false, true);
                     for (int i = 0; i < (int) tmp.size(); i++) {
-                        LOG("%6d -> '%s'\n", tmp[i], llama_token_to_piece(ctx, tmp[i]).c_str());
+                        LOG_INF("%6d -> '%s'\n", tmp[i], llama_token_to_piece(ctx, tmp[i]).c_str());
                     }
                 }
             }
         }
 
         if (params.input_prefix_bos) {
-            LOG("Input prefix with BOS\n");
+            LOG_INF("Input prefix with BOS\n");
         }
 
         if (!params.input_prefix.empty()) {
-            LOG("Input prefix: '%s'\n", params.input_prefix.c_str());
+            LOG_INF("Input prefix: '%s'\n", params.input_prefix.c_str());
             if (params.verbose_prompt) {
                 auto tmp = ::llama_tokenize(ctx, params.input_prefix, true, true);
                 for (int i = 0; i < (int) tmp.size(); i++) {
-                    LOG("%6d -> '%s'\n", tmp[i], llama_token_to_piece(ctx, tmp[i]).c_str());
+                    LOG_INF("%6d -> '%s'\n", tmp[i], llama_token_to_piece(ctx, tmp[i]).c_str());
                 }
             }
         }
 
         if (!params.input_suffix.empty()) {
-            LOG("Input suffix: '%s'\n", params.input_suffix.c_str());
+            LOG_INF("Input suffix: '%s'\n", params.input_suffix.c_str());
             if (params.verbose_prompt) {
                 auto tmp = ::llama_tokenize(ctx, params.input_suffix, false, true);
                 for (int i = 0; i < (int) tmp.size(); i++) {
-                    LOG("%6d -> '%s'\n", tmp[i], llama_token_to_piece(ctx, tmp[i]).c_str());
+                    LOG_INF("%6d -> '%s'\n", tmp[i], llama_token_to_piece(ctx, tmp[i]).c_str());
                 }
             }
         }
@@ -469,7 +474,7 @@ int main(int argc, char ** argv) {
       //GGML_ASSERT(n_ctx >= n_ctx_train * ga_n && "n_ctx must be at least n_ctx_train * grp_attn_n"); // NOLINT
         LOG_INF("self-extend: n_ctx_train = %d, grp_attn_n = %d, grp_attn_w = %d\n", n_ctx_train, ga_n, ga_w);
     }
-    LOG("\n");
+    LOG_INF("\n");
 
     if (params.interactive) {
         const char * control_message;
@@ -481,11 +486,11 @@ int main(int argc, char ** argv) {
                               " - To return control without starting a new line, end your input with '/'.\n"
                               " - If you want to submit another line, end your input with '\\'.\n";
         }
-        LOG("== Running in interactive mode. ==\n");
+        LOG_INF("== Running in interactive mode. ==\n");
 #if defined (__unix__) || (defined (__APPLE__) && defined (__MACH__)) || defined (_WIN32)
-        LOG(       " - Press Ctrl+C to interject at any time.\n");
+        LOG_INF(       " - Press Ctrl+C to interject at any time.\n");
 #endif
-        LOG(       "%s\n", control_message);
+        LOG_INF(       "%s\n", control_message);
 
         is_interacting = params.interactive_first;
     }

examples/perplexity/perplexity.cpp

@@ -444,7 +444,6 @@ static results_perplexity perplexity_v2(llama_context * ctx, const gpt_params &
             }
             LOG("%.2f minutes\n", total_seconds / 60.0);
         }
-        LOG("\n");
 
         //LOG_DBG("%s: using tokens %d...%d\n",__func__,params.n_ctx - params.ppl_stride + start, params.n_ctx + start);
         for (int j = n_ctx - params.ppl_stride - 1; j < n_ctx - 1; ++j) {
@@ -638,7 +637,6 @@ static results_perplexity perplexity(llama_context * ctx, const gpt_params & par
             }
             LOG("%.2f minutes\n", total_seconds / 60.0);
         }
-        LOG("\n");
 
         for (int seq = 0; seq < n_seq_batch; seq++) {
             const float * all_logits = num_batches > 1 ? logits.data() : llama_get_logits_ith(ctx, seq*n_ctx + first);
@@ -1961,6 +1959,7 @@ int main(int argc, char ** argv) {
 
     params.n_ctx = 512;
     params.logits_all = true;
+    params.escape = false;
 
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_PERPLEXITY)) {
         return 1;

examples/quantize/quantize.cpp

@@ -63,6 +63,16 @@ static const char * const LLM_KV_QUANTIZE_IMATRIX_DATASET    = "quantize.imatrix
 static const char * const LLM_KV_QUANTIZE_IMATRIX_N_ENTRIES  = "quantize.imatrix.entries_count";
 static const char * const LLM_KV_QUANTIZE_IMATRIX_N_CHUNKS   = "quantize.imatrix.chunks_count";
 
+static bool striequals(const char * a, const char * b) {
+    while (*a && *b) {
+        if (std::tolower(*a) != std::tolower(*b)) {
+            return false;
+        }
+        a++; b++;
+    }
+    return *a == *b;
+}
+
 static bool try_parse_ftype(const std::string & ftype_str_in, llama_ftype & ftype, std::string & ftype_str_out) {
     std::string ftype_str;
 
@@ -70,7 +80,7 @@ static bool try_parse_ftype(const std::string & ftype_str_in, llama_ftype & ftyp
         ftype_str.push_back(std::toupper(ch));
     }
     for (auto & it : QUANT_OPTIONS) {
-        if (it.name == ftype_str) {
+        if (striequals(it.name.c_str(), ftype_str.c_str())) {
             ftype = it.ftype;
             ftype_str_out = it.name;
             return true;
@@ -225,15 +235,15 @@ static int prepare_imatrix(const std::string & imatrix_file,
 }
 
 static ggml_type parse_ggml_type(const char * arg) {
-    ggml_type result = GGML_TYPE_COUNT;
+    for (int i = 0; i < GGML_TYPE_COUNT; ++i) {
-    for (int j = 0; j < GGML_TYPE_COUNT; ++j) {
+        auto type = (ggml_type)i;
-        auto type = ggml_type(j);
         const auto * name = ggml_type_name(type);
-        if (name && strcmp(arg, name) == 0) {
+        if (name && striequals(name, arg)) {
-            result = type; break;
+            return type;
         }
     }
-    return result;
+    fprintf(stderr, "%s: invalid ggml_type '%s'\n", __func__, arg);
+    return GGML_TYPE_COUNT;
 }
 
 int main(int argc, char ** argv) {
@@ -254,12 +264,18 @@ int main(int argc, char ** argv) {
         } else if (strcmp(argv[arg_idx], "--output-tensor-type") == 0) {
             if (arg_idx < argc-1) {
                 params.output_tensor_type = parse_ggml_type(argv[++arg_idx]);
+                if (params.output_tensor_type == GGML_TYPE_COUNT) {
+                    usage(argv[0]);
+                }
             } else {
                 usage(argv[0]);
             }
         } else if (strcmp(argv[arg_idx], "--token-embedding-type") == 0) {
             if (arg_idx < argc-1) {
                 params.token_embedding_type = parse_ggml_type(argv[++arg_idx]);
+                if (params.token_embedding_type == GGML_TYPE_COUNT) {
+                    usage(argv[0]);
+                }
             } else {
                 usage(argv[0]);
             }

examples/server/README.md

@@ -7,6 +7,7 @@ Set of LLM REST APIs and a simple web front end to interact with llama.cpp.
 **Features:**
  * LLM inference of F16 and quantized models on GPU and CPU
  * [OpenAI API](https://github.com/openai/openai-openapi) compatible chat completions and embeddings routes
+ * Reranking endoint (WIP: https://github.com/ggerganov/llama.cpp/pull/9510)
  * Parallel decoding with multi-user support
  * Continuous batching
  * Multimodal (wip)
@@ -17,12 +18,13 @@ The project is under active development, and we are [looking for feedback and co
 
 ## Usage
 
+**Common params**
+
 | Argument | Explanation |
 | -------- | ----------- |
 | `-h, --help, --usage` | print usage and exit |
 | `--version` | show version and build info |
-| `-v, --verbose` | print verbose information |
+| `--verbose-prompt` | print a verbose prompt before generation (default: false) |
-| `--verbosity N` | set specific verbosity level (default: 0) |
 | `-t, --threads N` | number of threads to use during generation (default: -1)<br/>(env: LLAMA_ARG_THREADS) |
 | `-tb, --threads-batch N` | number of threads to use during batch and prompt processing (default: same as --threads) |
 | `-C, --cpu-mask M` | CPU affinity mask: arbitrarily long hex. Complements cpu-range (default: "") |
@@ -42,13 +44,63 @@ The project is under active development, and we are [looking for feedback and co
 | `--keep N` | number of tokens to keep from the initial prompt (default: 0, -1 = all) |
 | `-fa, --flash-attn` | enable Flash Attention (default: disabled)<br/>(env: LLAMA_ARG_FLASH_ATTN) |
 | `-p, --prompt PROMPT` | prompt to start generation with |
+| `--no-perf` | disable internal libllama performance timings (default: false)<br/>(env: LLAMA_ARG_NO_PERF) |
 | `-f, --file FNAME` | a file containing the prompt (default: none) |
 | `-bf, --binary-file FNAME` | binary file containing the prompt (default: none) |
 | `-e, --escape` | process escapes sequences (\n, \r, \t, \', \", \\) (default: true) |
 | `--no-escape` | do not process escape sequences |
-| `--spm-infill` | use Suffix/Prefix/Middle pattern for infill (instead of Prefix/Suffix/Middle) as some models prefer this. (default: disabled) |
+| `--rope-scaling {none,linear,yarn}` | RoPE frequency scaling method, defaults to linear unless specified by the model<br/>(env: LLAMA_ARG_ROPE_SCALING_TYPE) |
+| `--rope-scale N` | RoPE context scaling factor, expands context by a factor of N<br/>(env: LLAMA_ARG_ROPE_SCALE) |
+| `--rope-freq-base N` | RoPE base frequency, used by NTK-aware scaling (default: loaded from model)<br/>(env: LLAMA_ARG_ROPE_FREQ_BASE) |
+| `--rope-freq-scale N` | RoPE frequency scaling factor, expands context by a factor of 1/N<br/>(env: LLAMA_ARG_ROPE_FREQ_SCALE) |
+| `--yarn-orig-ctx N` | YaRN: original context size of model (default: 0 = model training context size)<br/>(env: LLAMA_ARG_YARN_ORIG_CTX) |
+| `--yarn-ext-factor N` | YaRN: extrapolation mix factor (default: -1.0, 0.0 = full interpolation)<br/>(env: LLAMA_ARG_YARN_EXT_FACTOR) |
+| `--yarn-attn-factor N` | YaRN: scale sqrt(t) or attention magnitude (default: 1.0)<br/>(env: LLAMA_ARG_YARN_ATTN_FACTOR) |
+| `--yarn-beta-slow N` | YaRN: high correction dim or alpha (default: 1.0)<br/>(env: LLAMA_ARG_YARN_BETA_SLOW) |
+| `--yarn-beta-fast N` | YaRN: low correction dim or beta (default: 32.0)<br/>(env: LLAMA_ARG_YARN_BETA_FAST) |
+| `-gan, --grp-attn-n N` | group-attention factor (default: 1)<br/>(env: LLAMA_ARG_GRP_ATTN_N) |
+| `-gaw, --grp-attn-w N` | group-attention width (default: 512.0)<br/>(env: LLAMA_ARG_GRP_ATTN_W) |
+| `-dkvc, --dump-kv-cache` | verbose print of the KV cache |
+| `-nkvo, --no-kv-offload` | disable KV offload<br/>(env: LLAMA_ARG_NO_KV_OFFLOAD) |
+| `-ctk, --cache-type-k TYPE` | KV cache data type for K (default: f16)<br/>(env: LLAMA_ARG_CACHE_TYPE_K) |
+| `-ctv, --cache-type-v TYPE` | KV cache data type for V (default: f16)<br/>(env: LLAMA_ARG_CACHE_TYPE_V) |
+| `-dt, --defrag-thold N` | KV cache defragmentation threshold (default: -1.0, < 0 - disabled)<br/>(env: LLAMA_ARG_DEFRAG_THOLD) |
+| `-np, --parallel N` | number of parallel sequences to decode (default: 1)<br/>(env: LLAMA_ARG_N_PARALLEL) |
+| `--mlock` | force system to keep model in RAM rather than swapping or compressing<br/>(env: LLAMA_ARG_MLOCK) |
+| `--no-mmap` | do not memory-map model (slower load but may reduce pageouts if not using mlock)<br/>(env: LLAMA_ARG_NO_MMAP) |
+| `--numa TYPE` | attempt optimizations that help on some NUMA systems<br/>- distribute: spread execution evenly over all nodes<br/>- isolate: only spawn threads on CPUs on the node that execution started on<br/>- numactl: use the CPU map provided by numactl<br/>if run without this previously, it is recommended to drop the system page cache before using this<br/>see https://github.com/ggerganov/llama.cpp/issues/1437<br/>(env: LLAMA_ARG_NUMA) |
+| `-ngl, --gpu-layers, --n-gpu-layers N` | number of layers to store in VRAM<br/>(env: LLAMA_ARG_N_GPU_LAYERS) |
+| `-sm, --split-mode {none,layer,row}` | how to split the model across multiple GPUs, one of:<br/>- none: use one GPU only<br/>- layer (default): split layers and KV across GPUs<br/>- row: split rows across GPUs<br/>(env: LLAMA_ARG_SPLIT_MODE) |
+| `-ts, --tensor-split N0,N1,N2,...` | fraction of the model to offload to each GPU, comma-separated list of proportions, e.g. 3,1<br/>(env: LLAMA_ARG_TENSOR_SPLIT) |
+| `-mg, --main-gpu INDEX` | the GPU to use for the model (with split-mode = none), or for intermediate results and KV (with split-mode = row) (default: 0)<br/>(env: LLAMA_ARG_MAIN_GPU) |
+| `--check-tensors` | check model tensor data for invalid values (default: false) |
+| `--override-kv KEY=TYPE:VALUE` | advanced option to override model metadata by key. may be specified multiple times.<br/>types: int, float, bool, str. example: --override-kv tokenizer.ggml.add_bos_token=bool:false |
+| `--lora FNAME` | path to LoRA adapter (can be repeated to use multiple adapters) |
+| `--lora-scaled FNAME SCALE` | path to LoRA adapter with user defined scaling (can be repeated to use multiple adapters) |
+| `--control-vector FNAME` | add a control vector<br/>note: this argument can be repeated to add multiple control vectors |
+| `--control-vector-scaled FNAME SCALE` | add a control vector with user defined scaling SCALE<br/>note: this argument can be repeated to add multiple scaled control vectors |
+| `--control-vector-layer-range START END` | layer range to apply the control vector(s) to, start and end inclusive |
+| `-m, --model FNAME` | model path (default: `models/$filename` with filename from `--hf-file` or `--model-url` if set, otherwise models/7B/ggml-model-f16.gguf)<br/>(env: LLAMA_ARG_MODEL) |
+| `-mu, --model-url MODEL_URL` | model download url (default: unused)<br/>(env: LLAMA_ARG_MODEL_URL) |
+| `-hfr, --hf-repo REPO` | Hugging Face model repository (default: unused)<br/>(env: LLAMA_ARG_HF_REPO) |
+| `-hff, --hf-file FILE` | Hugging Face model file (default: unused)<br/>(env: LLAMA_ARG_HF_FILE) |
+| `-hft, --hf-token TOKEN` | Hugging Face access token (default: value from HF_TOKEN environment variable)<br/>(env: HF_TOKEN) |
+| `-ld, --logdir LOGDIR` | path under which to save YAML logs (no logging if unset) |
+| `--log-disable` | Log disable |
+| `--log-file FNAME` | Log to file |
+| `--log-colors` | Enable colored logging<br/>(env: LLAMA_LOG_COLORS) |
+| `-v, --verbose, --log-verbose` | Set verbosity level to infinity (i.e. log all messages, useful for debugging) |
+| `-lv, --verbosity, --log-verbosity N` | Set the verbosity threshold. Messages with a higher verbosity will be ignored.<br/>(env: LLAMA_LOG_VERBOSITY) |
+| `--log-prefix` | Enable prefx in log messages<br/>(env: LLAMA_LOG_PREFIX) |
+| `--log-timestamps` | Enable timestamps in log messages<br/>(env: LLAMA_LOG_TIMESTAMPS) |
+
+
+**Sampling params**
+
+| Argument | Explanation |
+| -------- | ----------- |
 | `--samplers SAMPLERS` | samplers that will be used for generation in the order, separated by ';'<br/>(default: top_k;tfs_z;typ_p;top_p;min_p;temperature) |
-| `-s, --seed SEED` | RNG seed (default: -1, use random seed for < 0) |
+| `-s, --seed SEED` | RNG seed (default: 4294967295, use random seed for 4294967295) |
 | `--sampling-seq SEQUENCE` | simplified sequence for samplers that will be used (default: kfypmt) |
 | `--ignore-eos` | ignore end of stream token and continue generating (implies --logit-bias EOS-inf) |
 | `--penalize-nl` | penalize newline tokens (default: false) |
@@ -71,54 +123,29 @@ The project is under active development, and we are [looking for feedback and co
 | `--grammar GRAMMAR` | BNF-like grammar to constrain generations (see samples in grammars/ dir) (default: '') |
 | `--grammar-file FNAME` | file to read grammar from |
 | `-j, --json-schema SCHEMA` | JSON schema to constrain generations (https://json-schema.org/), e.g. `{}` for any JSON object<br/>For schemas w/ external $refs, use --grammar + example/json_schema_to_grammar.py instead |
-| `--rope-scaling {none,linear,yarn}` | RoPE frequency scaling method, defaults to linear unless specified by the model |
+
-| `--rope-scale N` | RoPE context scaling factor, expands context by a factor of N |
+
-| `--rope-freq-base N` | RoPE base frequency, used by NTK-aware scaling (default: loaded from model) |
+**Example-specific params**
-| `--rope-freq-scale N` | RoPE frequency scaling factor, expands context by a factor of 1/N |
+
-| `--yarn-orig-ctx N` | YaRN: original context size of model (default: 0 = model training context size) |
+| Argument | Explanation |
-| `--yarn-ext-factor N` | YaRN: extrapolation mix factor (default: -1.0, 0.0 = full interpolation) |
+| -------- | ----------- |
-| `--yarn-attn-factor N` | YaRN: scale sqrt(t) or attention magnitude (default: 1.0) |
+| `--no-context-shift` | disables context shift on inifinite text generation (default: disabled)<br/>(env: LLAMA_ARG_NO_CONTEXT_SHIFT) |
-| `--yarn-beta-slow N` | YaRN: high correction dim or alpha (default: 1.0) |
+| `-sp, --special` | special tokens output enabled (default: false) |
-| `--yarn-beta-fast N` | YaRN: low correction dim or beta (default: 32.0) |
+| `--spm-infill` | use Suffix/Prefix/Middle pattern for infill (instead of Prefix/Suffix/Middle) as some models prefer this. (default: disabled) |
-| `-gan, --grp-attn-n N` | group-attention factor (default: 1) |
+| `--pooling {none,mean,cls,last,rank}` | pooling type for embeddings, use model default if unspecified<br/>(env: LLAMA_ARG_POOLING) |
-| `-gaw, --grp-attn-w N` | group-attention width (default: 512.0) |
-| `-dkvc, --dump-kv-cache` | verbose print of the KV cache |
-| `-nkvo, --no-kv-offload` | disable KV offload |
-| `-ctk, --cache-type-k TYPE` | KV cache data type for K (default: f16) |
-| `-ctv, --cache-type-v TYPE` | KV cache data type for V (default: f16) |
-| `-dt, --defrag-thold N` | KV cache defragmentation threshold (default: -1.0, < 0 - disabled)<br/>(env: LLAMA_ARG_DEFRAG_THOLD) |
-| `-np, --parallel N` | number of parallel sequences to decode (default: 1)<br/>(env:  LLAMA_ARG_N_PARALLEL) |
 | `-cb, --cont-batching` | enable continuous batching (a.k.a dynamic batching) (default: enabled)<br/>(env: LLAMA_ARG_CONT_BATCHING) |
 | `-nocb, --no-cont-batching` | disable continuous batching<br/>(env: LLAMA_ARG_NO_CONT_BATCHING) |
-| `--mlock` | force system to keep model in RAM rather than swapping or compressing |
+| `-a, --alias STRING` | set alias for model name (to be used by REST API)<br/>(env: LLAMA_ARG_ALIAS) |
-| `--no-mmap` | do not memory-map model (slower load but may reduce pageouts if not using mlock) |
-| `--numa TYPE` | attempt optimizations that help on some NUMA systems<br/>- distribute: spread execution evenly over all nodes<br/>- isolate: only spawn threads on CPUs on the node that execution started on<br/>- numactl: use the CPU map provided by numactl<br/>if run without this previously, it is recommended to drop the system page cache before using this<br/>see https://github.com/ggerganov/llama.cpp/issues/1437 |
-| `-ngl, --gpu-layers, --n-gpu-layers N` | number of layers to store in VRAM<br/>(env: LLAMA_ARG_N_GPU_LAYERS) |
-| `-sm, --split-mode {none,layer,row}` | how to split the model across multiple GPUs, one of:<br/>- none: use one GPU only<br/>- layer (default): split layers and KV across GPUs<br/>- row: split rows across GPUs |
-| `-ts, --tensor-split N0,N1,N2,...` | fraction of the model to offload to each GPU, comma-separated list of proportions, e.g. 3,1 |
-| `-mg, --main-gpu INDEX` | the GPU to use for the model (with split-mode = none), or for intermediate results and KV (with split-mode = row) (default: 0) |
-| `--check-tensors` | check model tensor data for invalid values (default: false) |
-| `--override-kv KEY=TYPE:VALUE` | advanced option to override model metadata by key. may be specified multiple times.<br/>types: int, float, bool, str. example: --override-kv tokenizer.ggml.add_bos_token=bool:false |
-| `--lora FNAME` | path to LoRA adapter (can be repeated to use multiple adapters) |
-| `--lora-scaled FNAME SCALE` | path to LoRA adapter with user defined scaling (can be repeated to use multiple adapters) |
-| `--control-vector FNAME` | add a control vector<br/>note: this argument can be repeated to add multiple control vectors |
-| `--control-vector-scaled FNAME SCALE` | add a control vector with user defined scaling SCALE<br/>note: this argument can be repeated to add multiple scaled control vectors |
-| `--control-vector-layer-range START END` | layer range to apply the control vector(s) to, start and end inclusive |
-| `-a, --alias STRING` | set alias for model name (to be used by REST API) |
-| `-m, --model FNAME` | model path (default: `models/$filename` with filename from `--hf-file` or `--model-url` if set, otherwise models/7B/ggml-model-f16.gguf)<br/>(env: LLAMA_ARG_MODEL) |
-| `-mu, --model-url MODEL_URL` | model download url (default: unused)<br/>(env: LLAMA_ARG_MODEL_URL) |
-| `-hfr, --hf-repo REPO` | Hugging Face model repository (default: unused)<br/>(env: LLAMA_ARG_HF_REPO) |
-| `-hff, --hf-file FILE` | Hugging Face model file (default: unused)<br/>(env: LLAMA_ARG_HF_FILE) |
-| `-hft, --hf-token TOKEN` | Hugging Face access token (default: value from HF_TOKEN environment variable)<br/>(env: HF_TOKEN) |
 | `--host HOST` | ip address to listen (default: 127.0.0.1)<br/>(env: LLAMA_ARG_HOST) |
 | `--port PORT` | port to listen (default: 8080)<br/>(env: LLAMA_ARG_PORT) |
-| `--path PATH` | path to serve static files from (default: ) |
+| `--path PATH` | path to serve static files from (default: )<br/>(env: LLAMA_ARG_STATIC_PATH) |
 | `--embedding, --embeddings` | restrict to only support embedding use case; use only with dedicated embedding models (default: disabled)<br/>(env: LLAMA_ARG_EMBEDDINGS) |
+| `--reranking, --rerank` | enable reranking endpoint on server (default: disabled)<br/>(env: LLAMA_ARG_RERANKING) |
 | `--api-key KEY` | API key to use for authentication (default: none)<br/>(env: LLAMA_API_KEY) |
 | `--api-key-file FNAME` | path to file containing API keys (default: none) |
-| `--ssl-key-file FNAME` | path to file a PEM-encoded SSL private key |
+| `--ssl-key-file FNAME` | path to file a PEM-encoded SSL private key<br/>(env: LLAMA_ARG_SSL_KEY_FILE) |
-| `--ssl-cert-file FNAME` | path to file a PEM-encoded SSL certificate |
+| `--ssl-cert-file FNAME` | path to file a PEM-encoded SSL certificate<br/>(env: LLAMA_ARG_SSL_CERT_FILE) |
-| `-to, --timeout N` | server read/write timeout in seconds (default: 600) |
+| `-to, --timeout N` | server read/write timeout in seconds (default: 600)<br/>(env: LLAMA_ARG_TIMEOUT) |
 | `--threads-http N` | number of threads used to process HTTP requests (default: -1)<br/>(env: LLAMA_ARG_THREADS_HTTP) |
 | `-spf, --system-prompt-file FNAME` | set a file to load a system prompt (initial prompt of all slots), this is useful for chat applications |
 | `--metrics` | enable prometheus compatible metrics endpoint (default: disabled)<br/>(env: LLAMA_ARG_ENDPOINT_METRICS) |
@@ -127,13 +154,7 @@ The project is under active development, and we are [looking for feedback and co
 | `--chat-template JINJA_TEMPLATE` | set custom jinja chat template (default: template taken from model's metadata)<br/>if suffix/prefix are specified, template will be disabled<br/>only commonly used templates are accepted:<br/>https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template<br/>(env: LLAMA_ARG_CHAT_TEMPLATE) |
 | `-sps, --slot-prompt-similarity SIMILARITY` | how much the prompt of a request must match the prompt of a slot in order to use that slot (default: 0.50, 0.0 = disabled)<br/> |
 | `--lora-init-without-apply` | load LoRA adapters without applying them (apply later via POST /lora-adapters) (default: disabled) |
-| `-ld, --logdir LOGDIR` | path under which to save YAML logs (no logging if unset) |
+
-| `--log-test` | Log test |
-| `--log-disable` | Log disable |
-| `--log-enable` | Log enable |
-| `--log-new` | Log new |
-| `--log-append` | Log append |
-| `--log-file FNAME` | Log file |
 
 Note: If both command line argument and environment variable are both set for the same param, the argument will take precedence over env var.
 
@@ -461,6 +482,39 @@ The same as [the embedding example](../embedding) does.
 
     `image_data`: An array of objects to hold base64-encoded image `data` and its `id`s to be reference in `content`. You can determine the place of the image in the content as in the following: `Image: [img-21].\nCaption: This is a picture of a house`. In this case, `[img-21]` will be replaced by the embeddings of the image with id `21` in the following `image_data` array: `{..., "image_data": [{"data": "<BASE64_STRING>", "id": 21}]}`. Use `image_data` only with multimodal models, e.g., LLaVA.
 
+### POST `/reranking`: Rerank documents according to a given query
+
+Similar to https://jina.ai/reranker/ but might change in the future.
+Requires a reranker model (such as [bge-reranker-v2-m3](https://huggingface.co/BAAI/bge-reranker-v2-m3)) and the `--embedding --pooling rank` options.
+
+    *Options:*
+
+    `query`: The query against which the documents will be ranked.
+
+    `documents`: An array strings representing the documents to be ranked.
+
+    *Aliases:*
+      - `/rerank`
+      - `/v1/rerank`
+      - `/v1/reranking`
+
+    *Examples:*
+
+    ```shell
+    curl http://127.0.0.1:8012/v1/rerank \
+        -H "Content-Type: application/json" \
+        -d '{
+            "model": "some-model",
+                "query": "What is panda?",
+                "top_n": 3,
+                "documents": [
+                    "hi",
+                "it is a bear",
+                "The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China."
+                ]
+        }' | jq
+    ```
+
 ### POST `/infill`: For code infilling.
 
 Takes a prefix and a suffix and returns the predicted completion as stream.

examples/server/server.cpp

@@ -92,6 +92,7 @@ enum server_task_type {
 enum server_task_cmpl_type {
     SERVER_TASK_CMPL_TYPE_NORMAL,
     SERVER_TASK_CMPL_TYPE_EMBEDDING,
+    SERVER_TASK_CMPL_TYPE_RERANK,
     SERVER_TASK_CMPL_TYPE_INFILL,
 };
 
@@ -172,6 +173,7 @@ struct server_slot {
     std::vector<completion_token_output> generated_token_probs;
 
     server_task_cmpl_type cmpl_type = SERVER_TASK_CMPL_TYPE_NORMAL;
+
     bool has_next_token = true;
     bool truncated      = false;
     bool stopped_eos    = false;
@@ -531,26 +533,38 @@ struct server_response {
 
     // add the id_task to the list of tasks waiting for response
     void add_waiting_task_id(int id_task) {
-        SRV_DBG("waiting for task id = %d\n", id_task);
+        SRV_DBG("add task %d to waiting list. current waiting = %d (before add)\n", id_task, (int) waiting_task_ids.size());
 
         std::unique_lock<std::mutex> lock(mutex_results);
         waiting_task_ids.insert(id_task);
     }
 
     void add_waiting_tasks(const std::vector<server_task> & tasks) {
-        for (const auto & t : tasks) {
+        std::unique_lock<std::mutex> lock(mutex_results);
-            add_waiting_task_id(t.id);
+
+        for (const auto & task : tasks) {
+            SRV_DBG("add task %d to waiting list. current waiting = %d (before add)\n", task.id, (int) waiting_task_ids.size());
+            waiting_task_ids.insert(task.id);
         }
     }
 
     // when the request is finished, we can remove task associated with it
     void remove_waiting_task_id(int id_task) {
-        SRV_DBG("task id = %d is done\n", id_task);
+        SRV_DBG("remove task %d from waiting list. current waiting = %d (before remove)\n", id_task, (int) waiting_task_ids.size());
 
         std::unique_lock<std::mutex> lock(mutex_results);
         waiting_task_ids.erase(id_task);
     }
 
+    void remove_waiting_task_ids(const std::unordered_set<int> & id_tasks) {
+        std::unique_lock<std::mutex> lock(mutex_results);
+
+        for (const auto & id_task : id_tasks) {
+            SRV_DBG("remove task %d from waiting list. current waiting = %d (before remove)\n", id_task, (int) waiting_task_ids.size());
+            waiting_task_ids.erase(id_task);
+        }
+    }
+
     // This function blocks the thread until there is a response for one of the id_tasks
     server_task_result recv(const std::unordered_set<int> & id_tasks) {
         while (true) {
@@ -942,8 +956,17 @@ struct server_context {
                 slot.prompt = *prompt;
             } else if (prompt->is_array() && prompt->size() == 1 && prompt->at(0).is_array()) {
                 slot.prompt = prompt->at(0);
+            } else if (prompt->is_array() && prompt->size() > 1) {
+                // array of strings
+                for (const auto & el : *prompt) {
+                    if (!el.is_string()) {
+                        send_error(task, "\"prompt\" must be a string, an array of strings or an array of integers", ERROR_TYPE_INVALID_REQUEST);
+                        return false;
+                    }
+                }
+                slot.prompt = *prompt;
             } else {
-                send_error(task, "\"prompt\" must be a string or an array of integers", ERROR_TYPE_INVALID_REQUEST);
+                send_error(task, "\"prompt\" must be a string, an array of strings or an array of integers", ERROR_TYPE_INVALID_REQUEST);
                 return false;
             }
         }
@@ -1168,6 +1191,15 @@ struct server_context {
             SLT_DBG(slot, "stopped by limit, n_decoded = %d, n_predict = %d\n", slot.n_decoded, slot.params.n_predict);
         }
 
+        // if context shift is disabled, we stop when it reaches the context limit
+        if (slot.n_decoded >= slot.n_ctx) {
+            slot.truncated      = true;
+            slot.stopped_limit  = true;
+            slot.has_next_token = false;
+
+            SLT_DBG(slot, "stopped due to running out of context capacity, n_decoded = %d, n_ctx = %d\n", slot.n_decoded, slot.n_ctx);
+        }
+
         if (llama_token_is_eog(model, result.tok)) {
             slot.stopped_eos    = true;
             slot.has_next_token = false;
@@ -1368,6 +1400,7 @@ struct server_context {
 
                 res.data = json {
                     {"embedding", std::vector<float>(n_embd, 0.0f)},
+                    {"index",     slot.index},
                 };
 
                 continue;
@@ -1386,6 +1419,44 @@ struct server_context {
         queue_results.send(res);
     }
 
+    void send_rerank(const server_slot & slot, const llama_batch & batch) {
+        server_task_result res;
+        res.id       = slot.id_task;
+        res.error    = false;
+        res.stop     = true;
+
+        for (int i = 0; i < batch.n_tokens; ++i) {
+            if (!batch.logits[i] || batch.seq_id[i][0] != slot.id + 1) {
+                continue;
+            }
+
+            const float * embd = llama_get_embeddings_seq(ctx, batch.seq_id[i][0]);
+            if (embd == NULL) {
+                embd = llama_get_embeddings_ith(ctx, i);
+            }
+
+            if (embd == NULL) {
+                SLT_ERR(slot, "failed to get embeddings, token = %d, seq_id = %d\n", batch.token[i], batch.seq_id[i][0]);
+
+                res.data = json {
+                    {"index", slot.index},
+                    {"score", -1e6},
+                };
+
+                continue;
+            }
+
+            res.data = json {
+                {"index", slot.index},
+                {"score", embd[0]},
+            };
+        }
+
+        SLT_DBG(slot, "sending rerank result, res = '%s'\n", res.data.dump().c_str());
+
+        queue_results.send(res);
+    }
+
     //
     // Functions to create new task(s) and receive result(s)
     //
@@ -1421,6 +1492,19 @@ struct server_context {
         // otherwise, it's a multiple-prompt task, we break it into smaller tasks
         else if (prompt.is_array()) {
             std::vector<json> prompts = prompt;
+            if (cmpl_type == SERVER_TASK_CMPL_TYPE_RERANK) {
+                // prompts[0] is the question
+                // the rest are the answers/documents
+                SRV_DBG("creating rerank tasks, n_prompts = %d\n", (int) prompts.size() - 1);
+                for (size_t i = 1; i < prompts.size(); i++) {
+                    json qd;
+                    qd.push_back(prompts[0]);
+                    qd.push_back(prompts[i]);
+                    data["index"] = i - 1;
+                    create_task(data, true, qd);
+                }
+            } else {
+                SRV_DBG("creating multi-prompt tasks, n_prompts = %d\n", (int) prompts.size());
                 for (size_t i = 0; i < prompts.size(); i++) {
                     const auto & e = prompts[i];
                     if (e.is_string() || json_is_array_of_numbers(e)) {
@@ -1431,6 +1515,7 @@ struct server_context {
                     }
                 }
             }
+        }
         // invalid case
         else {
             throw std::runtime_error(error_msg);
@@ -1468,10 +1553,12 @@ struct server_context {
             if (result.error) {
                 error_handler(result.data);
                 cancel_tasks(id_tasks);
-                break;
+                return;
             }
 
-            size_t idx = result.data["index"];
+            const size_t idx = result.data["index"];
+            GGML_ASSERT(idx < results.size() && "index out of range");
+
             results[idx] = result;
         }
         result_handler(results);
@@ -1815,6 +1902,14 @@ struct server_context {
         for (server_slot & slot : slots) {
             if (slot.ga_n == 1) {
                 if (slot.is_processing() && (int) system_tokens.size() + slot.n_past >= slot.n_ctx - 1) {
+                    if (!params.ctx_shift) {
+                        // this check is redundant (for good)
+                        // we should never get here, because generation should already stopped in process_token()
+                        slot.release();
+                        send_error(slot, "context shift is disabled", ERROR_TYPE_SERVER);
+                        continue;
+                    }
+
                     // Shift context
                     const int n_keep    = slot.params.n_keep + add_bos_token;
                     const int n_left    = (int) system_tokens.size() + slot.n_past - n_keep;
@@ -1874,6 +1969,7 @@ struct server_context {
         // track if this is an embedding or non-embedding batch
         // if we've added sampled tokens above, we are in non-embedding mode
         // -1: none, 0: non-embedding, 1: embedding
+        // TODO: make enum
         int32_t batch_type = batch.n_tokens > 0 ? 0 : -1;
 
         // next, batch any pending prompts without exceeding n_batch
@@ -1922,6 +2018,29 @@ struct server_context {
                             }
 
                             prompt_tokens = embd_inp;
+                        } else if (slot.cmpl_type == SERVER_TASK_CMPL_TYPE_RERANK) {
+                            // require slot.prompt to be array of 2 strings
+                            if (!slot.prompt.is_array() || slot.prompt.size() != 2) {
+                                SLT_ERR(slot, "%s", "invalid prompt for rerank task\n");
+                                slot.release();
+                                send_error(slot, "invalid prompt for rerank task", ERROR_TYPE_INVALID_REQUEST);
+                                continue;
+                            }
+
+                            // prompt: <s>query</s><s>doc</s>
+                            prompt_tokens.clear();
+                            prompt_tokens.push_back(llama_token_bos(model));
+                            {
+                                const auto part = tokenize(slot.prompt[0], false);
+                                prompt_tokens.insert(prompt_tokens.end(), part.begin(), part.end());
+                            }
+                            prompt_tokens.push_back(llama_token_eos(model));
+                            prompt_tokens.push_back(llama_token_bos(model));
+                            {
+                                const auto part = tokenize(slot.prompt[1], false);
+                                prompt_tokens.insert(prompt_tokens.end(), part.begin(), part.end());
+                            }
+                            prompt_tokens.push_back(llama_token_eos(model));
                         } else {
                             prompt_tokens = tokenize(slot.prompt, system_prompt.empty()); // add BOS if there isn't system prompt
                         }
@@ -1941,7 +2060,7 @@ struct server_context {
                             continue;
                         }
 
-                        if (slot.cmpl_type == SERVER_TASK_CMPL_TYPE_EMBEDDING) {
+                        if (slot.cmpl_type == SERVER_TASK_CMPL_TYPE_EMBEDDING || slot.cmpl_type == SERVER_TASK_CMPL_TYPE_RERANK) {
                             // this prompt is too large to process - discard it
                             if (slot.n_prompt_tokens > n_ubatch) {
                                 slot.release();
@@ -1949,6 +2068,14 @@ struct server_context {
                                 continue;
                             }
                         } else {
+                            if (!params.ctx_shift) {
+                                // if context shift is disabled, we make sure prompt size is smaller than KV size
+                                if ((int) system_tokens.size() + slot.n_prompt_tokens >= slot.n_ctx) {
+                                    slot.release();
+                                    send_error(slot, "the request exceeds the available context size. try increasing the context size or enable context shift", ERROR_TYPE_INVALID_REQUEST);
+                                    continue;
+                                }
+                            }
                             if (slot.params.n_keep < 0) {
                                 slot.params.n_keep = slot.n_prompt_tokens;
                             }
@@ -2011,7 +2138,8 @@ struct server_context {
                         slot.n_prompt_tokens_processed = 0;
                     }
 
-                    if (slot.cmpl_type == SERVER_TASK_CMPL_TYPE_EMBEDDING) {
+                    // non-causal tasks require to fit the entire prompt in the physical batch
+                    if (slot.cmpl_type == SERVER_TASK_CMPL_TYPE_EMBEDDING || slot.cmpl_type == SERVER_TASK_CMPL_TYPE_RERANK) {
                         // cannot fit the prompt in the current batch - will try next iter
                         if (batch.n_tokens + slot.n_prompt_tokens > n_batch) {
                             continue;
@@ -2019,7 +2147,10 @@ struct server_context {
                     }
 
                     // check that we are in the right batch_type, if not defer the slot
-                    bool slot_type = slot.cmpl_type == SERVER_TASK_CMPL_TYPE_EMBEDDING ? 1 : 0;
+                    const bool slot_type =
+                        slot.cmpl_type == SERVER_TASK_CMPL_TYPE_EMBEDDING ||
+                        slot.cmpl_type == SERVER_TASK_CMPL_TYPE_RERANK     ? 1 : 0;
+
                     if (batch_type == -1) {
                         batch_type = slot_type;
                     } else if (batch_type != slot_type) {
@@ -2192,6 +2323,13 @@ struct server_context {
                         continue; // continue loop of slots
                     }
 
+                    if (slot.cmpl_type == SERVER_TASK_CMPL_TYPE_RERANK) {
+                        send_rerank(slot, batch_view);
+                        slot.release();
+                        slot.i_batch = -1;
+                        continue; // continue loop of slots
+                    }
+
                     // prompt evaluated for next-token prediction
                     slot.state = SLOT_STATE_GENERATING;
                 } else if (slot.state != SLOT_STATE_GENERATING) {
@@ -2319,6 +2457,10 @@ int main(int argc, char ** argv) {
         svr.reset(new httplib::Server());
     }
 #else
+    if (params.ssl_file_key != "" && params.ssl_file_cert != "") {
+        LOG_ERR("Server is built without SSL support\n");
+        return 1;
+    }
     svr.reset(new httplib::Server());
 #endif
 
@@ -2746,8 +2888,8 @@ int main(int argc, char ** argv) {
     };
 
     const auto handle_completions_generic = [&ctx_server, &res_error, &res_ok](server_task_cmpl_type cmpl_type, json & data, httplib::Response & res) {
-        if (ctx_server.params.embedding) {
+        if (ctx_server.params.embedding || ctx_server.params.reranking) {
-            res_error(res, format_error_response("This server does not support completions. Start it without `--embeddings`", ERROR_TYPE_NOT_SUPPORTED));
+            res_error(res, format_error_response("This server does not support completions. Start it without `--embeddings` or `--reranking`", ERROR_TYPE_NOT_SUPPORTED));
             return;
         }
 
@@ -2774,6 +2916,8 @@ int main(int argc, char ** argv) {
             }, [&](const json & error_data) {
                 res_error(res, error_data);
             });
+
+            ctx_server.queue_results.remove_waiting_task_ids(task_ids);
         } else {
             const auto chunked_content_provider = [task_ids, &ctx_server](size_t, httplib::DataSink & sink) {
                 ctx_server.receive_cmpl_results_stream(task_ids, [&](const server_task_result & result) -> bool {
@@ -2784,7 +2928,12 @@ int main(int argc, char ** argv) {
                 sink.done();
                 return false;
             };
-            res.set_chunked_content_provider("text/event-stream", chunked_content_provider);
+
+            auto on_complete = [task_ids, &ctx_server] (bool) {
+                ctx_server.queue_results.remove_waiting_task_ids(task_ids);
+            };
+
+            res.set_chunked_content_provider("text/event-stream", chunked_content_provider, on_complete);
         }
     };
 
@@ -2800,8 +2949,8 @@ int main(int argc, char ** argv) {
 
     // TODO: maybe merge this function with "handle_completions_generic"
     const auto handle_chat_completions = [&ctx_server, &params, &res_error, &res_ok, verbose](const httplib::Request & req, httplib::Response & res) {
-        if (ctx_server.params.embedding) {
+        if (ctx_server.params.embedding || ctx_server.params.reranking) {
-            res_error(res, format_error_response("This server does not support completions. Start it without `--embeddings`", ERROR_TYPE_NOT_SUPPORTED));
+            res_error(res, format_error_response("This server does not support completions. Start it without `--embeddings` or `--reranking`", ERROR_TYPE_NOT_SUPPORTED));
             return;
         }
 
@@ -2823,6 +2972,8 @@ int main(int argc, char ** argv) {
             }, [&](const json & error_data) {
                 res_error(res, error_data);
             });
+
+            ctx_server.queue_results.remove_waiting_task_ids(task_ids);
         } else {
             const auto chunked_content_provider = [task_ids, &ctx_server, completion_id](size_t, httplib::DataSink & sink) {
                 ctx_server.receive_cmpl_results_stream(task_ids, [&](const server_task_result & result) -> bool {
@@ -2844,7 +2995,12 @@ int main(int argc, char ** argv) {
                 sink.done();
                 return true;
             };
-            res.set_chunked_content_provider("text/event-stream", chunked_content_provider);
+
+            auto on_complete = [task_ids, &ctx_server] (bool) {
+                ctx_server.queue_results.remove_waiting_task_ids(task_ids);
+            };
+
+            res.set_chunked_content_provider("text/event-stream", chunked_content_provider, on_complete);
         }
     };
 
@@ -2918,6 +3074,11 @@ int main(int argc, char ** argv) {
     };
 
     const auto handle_embeddings = [&ctx_server, &res_error, &res_ok](const httplib::Request & req, httplib::Response & res) {
+        // TODO: somehow clean up this checks in the future
+        if (!ctx_server.params.embedding || ctx_server.params.reranking) {
+            res_error(res, format_error_response("This server does not support embeddings. Start it with `--embeddings` and without `--reranking`", ERROR_TYPE_NOT_SUPPORTED));
+            return;
+        }
         const json body = json::parse(req.body);
         bool is_openai = false;
 
@@ -2953,6 +3114,8 @@ int main(int argc, char ** argv) {
                 res_error(res, error_data);
                 error = true;
             });
+
+            ctx_server.queue_results.remove_waiting_task_ids(task_ids);
         }
 
         if (error) {
@@ -2966,6 +3129,79 @@ int main(int argc, char ** argv) {
         res_ok(res, root);
     };
 
+    const auto handle_rerank = [&ctx_server, &res_error, &res_ok](const httplib::Request & req, httplib::Response & res) {
+        if (!ctx_server.params.reranking) {
+            res_error(res, format_error_response("This server does not support reranking. Start it with `--reranking`", ERROR_TYPE_NOT_SUPPORTED));
+            return;
+        }
+        const json body = json::parse(req.body);
+
+        // TODO: implement
+        //int top_n = 1;
+        //if (body.count("top_n") != 1) {
+        //    top_n = body.at("top_n");
+        //} else {
+        //    res_error(res, format_error_response("\"top_n\" must be provided", ERROR_TYPE_INVALID_REQUEST));
+        //    return;
+        //}
+
+        json query;
+        if (body.count("query") == 1) {
+            query = body.at("query");
+            if (!query.is_string()) {
+                res_error(res, format_error_response("\"query\" must be a string", ERROR_TYPE_INVALID_REQUEST));
+                return;
+            }
+        } else {
+            res_error(res, format_error_response("\"query\" must be provided", ERROR_TYPE_INVALID_REQUEST));
+            return;
+        }
+
+        std::vector<std::string> documents = json_value(body, "documents", std::vector<std::string>());
+        if (documents.empty()) {
+            res_error(res, format_error_response("\"documents\" must be a non-empty string array", ERROR_TYPE_INVALID_REQUEST));
+            return;
+        }
+
+        // construct prompt object: array of ["query", "doc0", "doc1", ...]
+        json prompt;
+        prompt.push_back(query);
+        for (const auto & doc : documents) {
+            prompt.push_back(doc);
+        }
+
+        LOG_DBG("rerank prompt: %s\n", prompt.dump().c_str());
+
+        // create and queue the task
+        json responses = json::array();
+        bool error = false;
+        {
+            std::vector<server_task> tasks = ctx_server.create_tasks_cmpl({{"prompt", prompt}}, SERVER_TASK_CMPL_TYPE_RERANK);
+            ctx_server.queue_results.add_waiting_tasks(tasks);
+            ctx_server.queue_tasks.post(tasks);
+
+            // get the result
+            std::unordered_set<int> task_ids = server_task::get_list_id(tasks);
+
+            ctx_server.receive_cmpl_results(task_ids, [&](std::vector<server_task_result> & results) {
+                for (const auto & res : results) {
+                    responses.push_back(res.data);
+                }
+            }, [&](const json & error_data) {
+                res_error(res, error_data);
+                error = true;
+            });
+        }
+
+        if (error) {
+            return;
+        }
+
+        // write JSON response
+        json root = format_response_rerank(body, responses);
+        res_ok(res, root);
+    };
+
     const auto handle_lora_adapters_list = [&](const httplib::Request &, httplib::Response & res) {
         json result = json::array();
         for (size_t i = 0; i < ctx_server.loras.size(); ++i) {
@@ -3062,6 +3298,10 @@ int main(int argc, char ** argv) {
     svr->Post("/embedding",           handle_embeddings); // legacy
     svr->Post("/embeddings",          handle_embeddings);
     svr->Post("/v1/embeddings",       handle_embeddings);
+    svr->Post("/rerank",              handle_rerank);
+    svr->Post("/reranking",           handle_rerank);
+    svr->Post("/v1/rerank",           handle_rerank);
+    svr->Post("/v1/reranking",        handle_rerank);
     svr->Post("/tokenize",            handle_tokenize);
     svr->Post("/detokenize",          handle_detokenize);
     // LoRA adapters hotswap
@@ -3126,7 +3366,7 @@ int main(int argc, char ** argv) {
     }
 
     // print sample chat example to make it clear which template is used
-    LOG_INF("%s: chat template, built_in: %d, chat_example: '%s\n'", __func__, params.chat_template.empty(), llama_chat_format_example(ctx_server.model, params.chat_template).c_str());
+    LOG_INF("%s: chat template, built_in: %d, chat_example: '%s'\n", __func__, params.chat_template.empty(), llama_chat_format_example(ctx_server.model, params.chat_template).c_str());
 
     ctx_server.queue_tasks.on_new_task(std::bind(
                 &server_context::process_single_task, &ctx_server, std::placeholders::_1));

examples/server/tests/features/ctx_shift.feature

@@ -0,0 +1,62 @@
+@llama.cpp
+@ctx_shift
+Feature: llama.cpp server
+
+  Background: Server startup
+    Given a server listening on localhost:8080
+    And   a model file tinyllamas/stories260K.gguf from HF repo ggml-org/models
+    And   a model file test-model.gguf
+    And   a model alias tinyllama-2
+    And   BOS token is 1
+    And   42 as server seed
+    And   256 KV cache size
+    And   32 as batch size
+    And   2 slots
+
+  Scenario: Inference with context shift
+    And   64 server max tokens to predict
+    Then  the server is starting
+    Then  the server is healthy
+    Given a prompt:
+    """
+    Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.
+    Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.
+    Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.
+    Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.
+    """
+    And   a completion request with no api error
+    Then  64 tokens are predicted matching fun|Annaks|popcorns|pictry|bowl
+    And   the completion is  truncated
+    And   109 prompt tokens are processed
+
+  Scenario Outline: Inference without context shift
+    And   <n_predict> server max tokens to predict
+    And   disable context shifting
+    Then  the server is starting
+    Then  the server is healthy
+    Given a prompt:
+    """
+    Hi how are you
+    """
+    And   a completion request with no api error
+    Then  <n_token_output> tokens are predicted matching twind|Anna
+    And   the completion is <truncated> truncated
+    And   8 prompt tokens are processed
+    Examples:
+      | n_predict | n_token_output | truncated |
+      | 64        | 64             | not       |
+      | -1        | 120            |           |
+
+  Scenario: Inference without context shift (expected error: prompt too long)
+    And   disable context shifting
+    Then  the server is starting
+    Then  the server is healthy
+    Given a prompt:
+    """
+    Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.
+    Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.
+    Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.
+    Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.
+    """
+    And   a completion request with 400 api error
+

examples/server/tests/features/embeddings.feature

@@ -10,12 +10,12 @@ Feature: llama.cpp server
     And   42 as server seed
     And   2 slots
     # the bert-bge-small model has context size of 512
-    # since the generated prompts are as big as the batch size, we need to set the batch size to 512
+    # since the generated prompts are as big as the batch size, we need to set the batch size to <= 512
     # ref: https://huggingface.co/BAAI/bge-small-en-v1.5/blob/5c38ec7c405ec4b44b94cc5a9bb96e735b38267a/config.json#L20
-    And   512 as batch size
+    And   128 as batch size
-    And   512 as ubatch size
+    And   128 as ubatch size
-    And   2048 KV cache size
+    And   512 KV cache size
-    And   embeddings extraction
+    And   enable embeddings endpoint
     Then  the server is starting
     Then  the server is healthy
 
@@ -26,6 +26,20 @@ Feature: llama.cpp server
     """
     Then embeddings are generated
 
+  Scenario: Embedding (error: prompt too long)
+    When embeddings are computed for:
+    """
+    Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.
+    Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.
+    Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.
+    Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.
+    Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.
+    Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.
+    Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.
+    Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.
+    """
+    And  embeddings request with 500 api error
+
   Scenario: OAI Embeddings compatibility
     Given a model bert-bge-small
     When an OAI compatible embeddings computation request for:

examples/server/tests/features/rerank.feature

@@ -0,0 +1,42 @@
+@llama.cpp
+@rerank
+Feature: llama.cpp server
+
+  Background: Server startup
+    Given a server listening on localhost:8080
+    And   a model url https://huggingface.co/ggml-org/models/resolve/main/jina-reranker-v1-tiny-en/ggml-model-f16.gguf
+    And   a model file jina-reranker-v1-tiny-en.gguf
+    And   a model alias jina-reranker-v1-tiny-en
+    And   42 as server seed
+    And   2 slots
+    And   512 as batch size
+    And   512 as ubatch size
+    And   512 KV cache size
+    And   enable reranking endpoint
+    Then  the server is starting
+    Then  the server is healthy
+
+  Scenario: Rerank
+    Given a rerank query:
+      """
+      Machine learning is
+      """
+    And   a rerank document:
+      """
+      A machine is a physical system that uses power to apply forces and control movement to perform an action. The term is commonly applied to artificial devices, such as those employing engines or motors, but also to natural biological macromolecules, such as molecular machines.
+      """
+    And   a rerank document:
+      """
+      Learning is the process of acquiring new understanding, knowledge, behaviors, skills, values, attitudes, and preferences. The ability to learn is possessed by humans, non-human animals, and some machines; there is also evidence for some kind of learning in certain plants.
+      """
+    And   a rerank document:
+      """
+      Machine learning is a field of study in artificial intelligence concerned with the development and study of statistical algorithms that can learn from data and generalize to unseen data, and thus perform tasks without explicit instructions.
+      """
+    And   a rerank document:
+      """
+      Paris, capitale de la France, est une grande ville européenne et un centre mondial de l'art, de la mode, de la gastronomie et de la culture. Son paysage urbain du XIXe siècle est traversé par de larges boulevards et la Seine.
+      """
+    When  reranking request
+    Then  reranking results are returned
+    Then  reranking highest score is index 2 and lowest score is index 3

examples/server/tests/features/steps/steps.py

@@ -68,6 +68,7 @@ def step_server_config(context, server_fqdn: str, server_port: str):
     context.server_api_key = None
     context.server_continuous_batching = False
     context.server_embeddings = False
+    context.server_reranking = False
     context.server_metrics = False
     context.server_process = None
     context.seed = None
@@ -77,11 +78,16 @@ def step_server_config(context, server_fqdn: str, server_port: str):
     context.response_format = None
     context.temperature = None
     context.lora_file = None
+    context.disable_ctx_shift = False
 
     context.tasks_result = []
     context.concurrent_tasks = []
     context.prompts = []
 
+    context.reranking_query = None
+    context.reranking_documents = []
+    context.reranking_results = None
+
 
 @step('a model file {hf_file} from HF repo {hf_repo}')
 def step_download_hf_model(context, hf_file: str, hf_repo: str):
@@ -148,7 +154,7 @@ def step_n_slots(context, n_slots: int):
 
 @step('{n_predict:d} server max tokens to predict')
 def step_server_n_predict(context, n_predict: int):
-    context.n_server_predict = n_predict
+    context.n_server_predict = n_predict if n_predict > 0 else None
 
 
 @step('{slot_save_path} as slot save path')
@@ -171,15 +177,21 @@ def step_server_continuous_batching(context):
     context.server_continuous_batching = True
 
 
-@step('embeddings extraction')
+@step('enable embeddings endpoint')
 def step_server_embeddings(context):
     context.server_embeddings = True
 
+@step('enable reranking endpoint')
+def step_server_reranking(context):
+    context.server_reranking = True
 
 @step('prometheus compatible metrics exposed')
 def step_server_metrics(context):
     context.server_metrics = True
 
+@step('disable context shifting')
+def step_server_disable_ctx_shift(context):
+    context.disable_ctx_shift = True
 
 @step("the server is starting")
 def step_start_server(context):
@@ -257,7 +269,7 @@ async def step_all_slots_status(context, expected_slot_status_string: Literal['i
 @step('a completion request with {api_error} api error')
 @async_run_until_complete
 async def step_request_completion(context, api_error: Literal['raised'] | str):
-    expect_api_error = api_error == 'raised'
+    expect_api_error = api_error == 'raised' or api_error != 'no'
     seeds = await completions_seed(context, num_seeds=1)
     completion = await request_completion(context.prompts.pop(),
                                           seeds[0] if seeds is not None else seeds,
@@ -272,8 +284,11 @@ async def step_request_completion(context, api_error: Literal['raised'] | str):
     context.tasks_result.append(completion)
     if context.debug:
         print(f"Completion response: {completion}")
-    if expect_api_error:
+    if api_error == 'raised':
         assert completion == 401, f"completion must be an 401 status code: {completion}"
+    elif api_error.isdigit():
+        api_error_code = int(api_error)
+        assert completion == api_error_code, f"completion must be an {api_error_code} status code: {completion}"
 
 
 @step('{predicted_n:d} tokens are predicted matching {re_content}')
@@ -445,6 +460,14 @@ def step_impl(context, n_ga_w):
 def step_prompt_passkey(context):
     context.prompt_passkey = context_text(context)
 
+@step('a rerank query')
+def step_set_rerank_query(context):
+    context.reranking_query = context_text(context)
+    context.reranking_documents = []
+
+@step('a rerank document')
+def step_set_rerank_document(context):
+    context.reranking_documents.append(context_text(context))
 
 @step('{n_prompts:d} fixed prompts')
 def step_fixed_prompts(context, n_prompts):
@@ -612,6 +635,22 @@ async def step_compute_embedding(context):
     context.embeddings = await request_embedding(context_text(context), None, base_url=context.base_url)
 
 
+@step('reranking request')
+@async_run_until_complete
+async def step_compute_reranking(context):
+    async with aiohttp.ClientSession(timeout=DEFAULT_TIMEOUT_SECONDS) as session:
+        async with session.post(f'{context.base_url}/reranking',
+                                json={
+                                    "query": context.reranking_query,
+                                    "documents": context.reranking_documents,
+                                }) as response:
+            if response.status == 200:
+                response_json = await response.json()
+                context.reranking_results = response_json['results']
+            else:
+                context.reranking_results = response.status
+
+
 @step('all embeddings are the same')
 @async_run_until_complete
 async def step_all_embeddings_are_the_same(context):
@@ -645,6 +684,9 @@ def step_assert_embeddings(context):
     for embedding in context.embeddings:
         assert_embeddings(embedding)
 
+@step('embeddings request with {api_error_code:d} api error')
+def step_assert_embeddings(context, api_error_code: int):
+    assert context.embeddings == api_error_code, f"embeddings request must return code {api_error_code}, but got {context.embeddings}"
 
 @step('an OAI compatible embeddings computation request for')
 @async_run_until_complete
@@ -694,6 +736,24 @@ async def all_embeddings_are_generated(context):
     for i in range(n_embedding_requests):
         assert_embeddings(context.tasks_result.pop().pop())
 
+@step('reranking results are returned')
+def reranking_results_are_returned(context):
+    assert len(context.reranking_results) == len(context.reranking_documents)
+
+@step('reranking highest score is index {idx_high:d} and lowest score is index {idx_low:d}')
+def reranking_results_are_returned(context, idx_high: int, idx_low: int):
+    max_score, max_idx = 0, 0
+    min_score, min_idx = 0, 0
+    for res in context.reranking_results:
+        if max_score < res['relevance_score']:
+            max_score = res['relevance_score']
+            max_idx   = res['index']
+        if min_score > res['relevance_score']:
+            min_score = res['relevance_score']
+            min_idx   = res['index']
+    print(context.reranking_results)
+    assert max_idx == idx_high
+    assert min_idx == idx_low
 
 @step('adding special tokens')
 def step_tokenize_set_add_special(context):
@@ -1089,15 +1149,17 @@ async def oai_chat_completions(user_prompt,
     return completion_response
 
 
-async def request_embedding(content, seed, base_url=None) -> list[list[float]]:
+async def request_embedding(content, seed, base_url=None) -> list[list[float]] | int:
     async with aiohttp.ClientSession(timeout=DEFAULT_TIMEOUT_SECONDS) as session:
         async with session.post(f'{base_url}/embedding',
                                 json={
                                     "content": content,
                                 }) as response:
-            assert response.status == 200
+            if response.status == 200:
                 response_json = await response.json()
                 return [response_json['embedding']]
+            else:
+                return response.status
 
 
 async def request_oai_embeddings(input, seed,
@@ -1350,6 +1412,8 @@ def start_server_background(context):
         server_args.append('--cont-batching')
     if context.server_embeddings:
         server_args.append('--embedding')
+    if context.server_reranking:
+        server_args.append('--reranking')
     if context.server_metrics:
         server_args.append('--metrics')
     if context.model_alias:
@@ -1372,6 +1436,8 @@ def start_server_background(context):
         server_args.append('--verbose')
     if context.lora_file:
         server_args.extend(['--lora', context.lora_file])
+    if context.disable_ctx_shift:
+        server_args.extend(['--no-context-shift'])
 
     args = [str(arg) for arg in [context.server_path, *server_args]]
     print(f"bench: starting server with: {' '.join(args)}")

examples/server/tests/requirements.txt

@@ -1,6 +1,6 @@
 aiohttp~=3.9.3
 behave~=1.2.6
-huggingface_hub~=0.20.3
+huggingface_hub~=0.23.2
 numpy~=1.26.4
 openai~=1.30.3
 prometheus-client~=0.20.0

examples/server/utils.hpp

@@ -537,7 +537,7 @@ static json format_embeddings_response_oaicompat(const json & request, const jso
     json res = json {
         {"model", json_value(request, "model", std::string(DEFAULT_OAICOMPAT_MODEL))},
         {"object", "list"},
-        {"usage", json {
+        {"usage", json { // TODO: fill
             {"prompt_tokens", 0},
             {"total_tokens", 0}
         }},
@@ -547,6 +547,29 @@ static json format_embeddings_response_oaicompat(const json & request, const jso
     return res;
 }
 
+static json format_response_rerank(const json & request, const json & ranks) {
+    json data = json::array();
+    int i = 0;
+    for (const auto & rank : ranks) {
+        data.push_back(json{
+            {"index",    i++},
+            {"relevance_score", json_value(rank, "score", 0.0)},
+        });
+    }
+
+    json res = json {
+        {"model", json_value(request, "model", std::string(DEFAULT_OAICOMPAT_MODEL))},
+        {"object", "list"},
+        {"usage", json { // TODO: fill
+            {"prompt_tokens", 0},
+            {"total_tokens", 0}
+        }},
+        {"results", data}
+    };
+
+    return res;
+}
+
 static bool is_valid_utf8(const std::string & str) {
     const unsigned char* bytes = reinterpret_cast<const unsigned char*>(str.data());
     const unsigned char* end = bytes + str.length();

examples/speculative/speculative.cpp

@@ -32,6 +32,9 @@ struct seq_draft {
 int main(int argc, char ** argv) {
     gpt_params params;
 
+    // needed to get candidate probs even for temp <= 0.0
+    params.sparams.n_probs = 128;
+
     if (!gpt_params_parse(argc, argv, params, LLAMA_EXAMPLE_SPECULATIVE)) {
         return 1;
     }
@@ -49,7 +52,7 @@ int main(int argc, char ** argv) {
     // probability threshold for splitting a draft branch (only for n_seq_dft > 1)
     const float p_split  = params.p_split;
 
-    std::default_random_engine rng(params.sparams.seed);
+    std::default_random_engine rng(params.sparams.seed == LLAMA_DEFAULT_SEED ? std::random_device()() : params.sparams.seed);
     std::uniform_real_distribution<> u_dist;
 
     // init llama.cpp

flake.lock

@@ -20,11 +20,11 @@
     },
     "nixpkgs": {
       "locked": {
-        "lastModified": 1726062873,
+        "lastModified": 1727348695,
-        "narHash": "sha256-IiA3jfbR7K/B5+9byVi9BZGWTD4VSbWe8VLpp9B/iYk=",
+        "narHash": "sha256-J+PeFKSDV+pHL7ukkfpVzCOO7mBSrrpJ3svwBFABbhI=",
         "owner": "NixOS",
         "repo": "nixpkgs",
-        "rev": "4f807e8940284ad7925ebd0a0993d2a1791acb2f",
+        "rev": "1925c603f17fc89f4c8f6bf6f631a802ad85d784",
         "type": "github"
       },
       "original": {

ggml/include/ggml-backend.h

@@ -66,6 +66,7 @@ extern "C" {
     // "offset" refers to the offset of the tensor data for setting/getting data
     GGML_API GGML_CALL void ggml_backend_tensor_set(      struct ggml_tensor * tensor, const void * data, size_t offset, size_t size);
     GGML_API GGML_CALL void ggml_backend_tensor_get(const struct ggml_tensor * tensor,       void * data, size_t offset, size_t size);
+    GGML_API GGML_CALL void ggml_backend_tensor_memset(   struct ggml_tensor * tensor,     uint8_t value, size_t offset, size_t size);
 
     GGML_API void ggml_backend_synchronize(ggml_backend_t backend);
 
@@ -122,7 +123,7 @@ extern "C" {
     // The backend registry is a registry of all the available backends, and allows initializing backends in a generic way
 
     GGML_API size_t                     ggml_backend_reg_get_count(void);
-    GGML_API size_t                     ggml_backend_reg_find_by_name(const char * name);
+    GGML_API size_t                     ggml_backend_reg_find_by_name(const char * name); // returns index of backend with name, or SIZE_MAX if not found
     GGML_API ggml_backend_t             ggml_backend_reg_init_backend_from_str(const char * backend_str); // str is backend_name:params (params is optional)
     GGML_API const char *               ggml_backend_reg_get_name(size_t i);
     GGML_API ggml_backend_t             ggml_backend_reg_init_backend(size_t i, const char * params); // params is backend-specific

ggml/include/ggml-metal.h

@@ -25,9 +25,6 @@
 #include <stddef.h>
 #include <stdbool.h>
 
-// max memory buffers that can be mapped to the device
-#define GGML_METAL_MAX_BUFFERS 64
-
 struct ggml_tensor;
 struct ggml_cgraph;
 
@@ -48,8 +45,6 @@ GGML_API bool ggml_backend_is_metal(ggml_backend_t backend);
 
 GGML_API GGML_CALL ggml_backend_buffer_t ggml_backend_metal_buffer_from_ptr(void * data, size_t size, size_t max_size);
 
-GGML_API void ggml_backend_metal_set_n_cb(ggml_backend_t backend, int n_cb);
-
 GGML_API void ggml_backend_metal_set_abort_callback(ggml_backend_t backend, ggml_abort_callback abort_callback, void * user_data);
 
 GGML_API GGML_CALL ggml_backend_buffer_type_t ggml_backend_metal_buffer_type(void);

ggml/include/ggml.h

@@ -229,14 +229,16 @@
 #define GGML_MAX_PARAMS         2048
 #define GGML_MAX_CONTEXTS       64
 #define GGML_MAX_SRC            10
-#ifndef GGML_MAX_NAME
-#define GGML_MAX_NAME           64
 #define GGML_MAX_N_THREADS      512
-
-#endif
 #define GGML_MAX_OP_PARAMS      64
+
+#ifndef GGML_MAX_NAME
+#   define GGML_MAX_NAME        64
+#endif
+
 #define GGML_DEFAULT_N_THREADS  4
 #define GGML_DEFAULT_GRAPH_SIZE 2048
+
 #if UINTPTR_MAX == 0xFFFFFFFF
     #define GGML_MEM_ALIGN 4
 #else
@@ -259,21 +261,21 @@
 #define GGML_PAD(x, n) (((x) + (n) - 1) & ~((n) - 1))
 
 #ifndef NDEBUG
-#define GGML_UNREACHABLE() do { fprintf(stderr, "statement should be unreachable\n"); abort(); } while(0)
+#   define GGML_UNREACHABLE() do { fprintf(stderr, "statement should be unreachable\n"); abort(); } while(0)
 #elif defined(__GNUC__)
-#define GGML_UNREACHABLE() __builtin_unreachable()
+#   define GGML_UNREACHABLE() __builtin_unreachable()
 #elif defined(_MSC_VER)
-#define GGML_UNREACHABLE() __assume(0)
+#   define GGML_UNREACHABLE() __assume(0)
 #else
-#define GGML_UNREACHABLE() ((void) 0)
+#   define GGML_UNREACHABLE() ((void) 0)
 #endif
 
 #ifdef __cplusplus
-#define GGML_NORETURN [[noreturn]]
+#   define GGML_NORETURN [[noreturn]]
 #elif defined(_MSC_VER)
-#define GGML_NORETURN __declspec(noreturn)
+#   define GGML_NORETURN __declspec(noreturn)
 #else
-#define GGML_NORETURN _Noreturn
+#   define GGML_NORETURN _Noreturn
 #endif
 
 #define GGML_ABORT(...) ggml_abort(__FILE__, __LINE__, __VA_ARGS__)
@@ -534,6 +536,7 @@ extern "C" {
 
         GGML_OP_CROSS_ENTROPY_LOSS,
         GGML_OP_CROSS_ENTROPY_LOSS_BACK,
+        GGML_OP_OPT_STEP_ADAMW,
 
         GGML_OP_COUNT,
     };
@@ -569,12 +572,15 @@ extern "C" {
         GGML_LOG_LEVEL_WARN  = 2,
         GGML_LOG_LEVEL_ERROR = 3,
         GGML_LOG_LEVEL_DEBUG = 4,
+        GGML_LOG_LEVEL_CONT  = 5, // continue previous log
     };
 
+    // this tensor...
     enum ggml_tensor_flag {
-        GGML_TENSOR_FLAG_INPUT  = 1,
+        GGML_TENSOR_FLAG_INPUT  =  1, // ...is an input for the GGML compute graph
-        GGML_TENSOR_FLAG_OUTPUT = 2,
+        GGML_TENSOR_FLAG_OUTPUT =  2, // ...is an output for the GGML compute graph
-        GGML_TENSOR_FLAG_PARAM  = 4,
+        GGML_TENSOR_FLAG_PARAM  =  4, // ...contains trainable parameters
+        GGML_TENSOR_FLAG_LOSS   =  8, // ...defines loss for numerical optimization (multiple loss tensors add up)
     };
 
     // n-dimensional tensor
@@ -1404,14 +1410,14 @@ extern "C" {
     // supports 3D: a->ne[2] == b->ne[1]
     GGML_API struct ggml_tensor * ggml_get_rows(
             struct ggml_context * ctx,
-            struct ggml_tensor  * a,
+            struct ggml_tensor  * a,  // data
-            struct ggml_tensor  * b);
+            struct ggml_tensor  * b); // row indices
 
     GGML_API struct ggml_tensor * ggml_get_rows_back(
             struct ggml_context * ctx,
-            struct ggml_tensor  * a,
+            struct ggml_tensor  * a,  // gradients of ggml_get_rows result
-            struct ggml_tensor  * b,
+            struct ggml_tensor  * b,  // row indices
-            struct ggml_tensor  * c);
+            struct ggml_tensor  * c); // data for ggml_get_rows, only used for its shape
 
     GGML_API struct ggml_tensor * ggml_diag(
         struct ggml_context     * ctx,
@@ -1562,9 +1568,9 @@ extern "C" {
     // a - dy
     GGML_API struct ggml_tensor * ggml_rope_back(
             struct ggml_context * ctx,
-            struct ggml_tensor  * a,
+            struct ggml_tensor  * a, // gradients of ggml_rope result
-            struct ggml_tensor  * b,
+            struct ggml_tensor  * b, // positions
-            struct ggml_tensor  * c,
+            struct ggml_tensor  * c, // freq factors
             int                   n_dims,
             int                   mode,
             int                   n_ctx_orig,
@@ -1978,6 +1984,9 @@ extern "C" {
     typedef void (*ggml_custom2_op_t)(struct ggml_tensor * dst , const struct ggml_tensor * a, const struct ggml_tensor * b, int ith, int nth, void * userdata);
     typedef void (*ggml_custom3_op_t)(struct ggml_tensor * dst , const struct ggml_tensor * a, const struct ggml_tensor * b, const struct ggml_tensor * c, int ith, int nth, void * userdata);
 
+#define GGML_N_TASKS_MAX (-1)
+    // n_tasks == GGML_N_TASKS_MAX means to use max number of tasks
+
     GGML_API struct ggml_tensor * ggml_map_custom1(
             struct ggml_context   * ctx,
             struct ggml_tensor    * a,
@@ -2030,32 +2039,54 @@ extern "C" {
 
     GGML_API struct ggml_tensor * ggml_cross_entropy_loss(
             struct ggml_context * ctx,
-            struct ggml_tensor          * a,
+            struct ggml_tensor  * a,  // logits
-            struct ggml_tensor          * b);
+            struct ggml_tensor  * b); // labels
 
     GGML_API struct ggml_tensor * ggml_cross_entropy_loss_back(
+            struct ggml_context * ctx,
+            struct ggml_tensor  * a,  // logits
+            struct ggml_tensor  * b,  // labels
+            struct ggml_tensor  * c); // gradients of cross_entropy_loss result
+
+    // AdamW optimizer step
+    // Paper: https://arxiv.org/pdf/1711.05101v3.pdf
+    // PyTorch: https://pytorch.org/docs/stable/generated/torch.optim.AdamW.html
+    GGML_API struct ggml_tensor * ggml_opt_step_adamw(
             struct ggml_context * ctx,
             struct ggml_tensor  * a,
-            struct ggml_tensor          * b,
+            struct ggml_tensor  * grad,
-            struct ggml_tensor          * c);
+            float                 alpha,
+            float                 beta1,
+            float                 beta2,
+            float                 eps,
+            float                 wd); // weight decay
 
     //
     // automatic differentiation
     //
 
-    GGML_API void ggml_set_param(
+    GGML_API void ggml_set_param(struct ggml_context * ctx, struct ggml_tensor * tensor);
-            struct ggml_context * ctx,
+    GGML_API void ggml_set_loss(struct ggml_tensor * tensor);
-            struct ggml_tensor  * tensor);
 
     GGML_API void ggml_build_forward_expand (struct ggml_cgraph * cgraph, struct ggml_tensor * tensor);
-    GGML_API void ggml_build_backward_expand(struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, bool keep);
+    GGML_API void ggml_build_backward_expand(struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, bool accumulate);
+
+    GGML_API void ggml_build_opt_adamw(
+            struct ggml_context * ctx,
+            struct ggml_cgraph  * gf,
+            struct ggml_cgraph  * gb,
+            float                 alpha,
+            float                 beta1,
+            float                 beta2,
+            float                 eps,
+            float                 wd); // weight decay
 
     // graph allocation in a context
     GGML_API struct ggml_cgraph * ggml_new_graph       (struct ggml_context * ctx); // size = GGML_DEFAULT_GRAPH_SIZE, grads = false
     GGML_API struct ggml_cgraph * ggml_new_graph_custom(struct ggml_context * ctx, size_t size, bool grads);
     GGML_API struct ggml_cgraph * ggml_graph_dup       (struct ggml_context * ctx, struct ggml_cgraph * cgraph);
     GGML_API void                 ggml_graph_cpy       (struct ggml_cgraph * src, struct ggml_cgraph * dst);
-    GGML_API void                 ggml_graph_reset     (struct ggml_cgraph * cgraph);  // zero grads
+    GGML_API void                 ggml_graph_reset     (struct ggml_cgraph * cgraph); // set regular grads + optimizer momenta to 0, set loss grad to 1
     GGML_API void                 ggml_graph_clear     (struct ggml_cgraph * cgraph);
 
     GGML_API int                   ggml_graph_size   (struct ggml_cgraph * cgraph);
@@ -2481,6 +2512,9 @@ extern "C" {
     GGML_API int ggml_cpu_has_cann       (void);
     GGML_API int ggml_cpu_has_llamafile  (void);
 
+    // get the sve vector length in bytes
+    GGML_API int ggml_cpu_get_sve_cnt(void);
+
     //
     // Internal types and functions exposed for tests and benchmarks
     //

ggml/src/CMakeLists.txt

@@ -364,7 +364,7 @@ if (GGML_CUDA)
         if (GGML_MUSA)
             set_source_files_properties(${GGML_SOURCES_CUDA} PROPERTIES LANGUAGE CXX)
             foreach(SOURCE ${GGML_SOURCES_CUDA})
-                set_property(SOURCE ${SOURCE} PROPERTY COMPILE_FLAGS "-x musa -mtgpu --cuda-gpu-arch=mp_22")
+                set_property(SOURCE ${SOURCE} PROPERTY COMPILE_FLAGS "-x musa -mtgpu --cuda-gpu-arch=mp_21 --cuda-gpu-arch=mp_22")
             endforeach()
         endif()
 
@@ -1186,6 +1186,7 @@ elseif (CMAKE_OSX_ARCHITECTURES STREQUAL "x86_64" OR CMAKE_GENERATOR_PLATFORM_LW
         endif()
         if (GGML_AVX512)
             list(APPEND ARCH_FLAGS -mavx512f)
+            list(APPEND ARCH_FLAGS -mavx512dq)
             list(APPEND ARCH_FLAGS -mavx512bw)
         endif()
         if (GGML_AVX512_VBMI)

ggml/src/ggml-aarch64.c

@@ -1,4 +1,7 @@
-// SPDX-FileCopyrightText: Copyright 2024 Arm Ltd.
+// SPDX-FileCopyrightText: Copyright 2024 Arm Limited and/or its affiliates <open-source-office@arm.com>
+// SPDX-License-Identifier: MIT
+//
+
 #define GGML_COMMON_IMPL_C
 #include "ggml-common.h"
 
@@ -39,11 +42,44 @@
 //
 #if defined(__AVX__)
 #if defined(__F16C__)
+#if defined(__AVX512F__)
+#define GGML_F32Cx8x2_LOAD(x, y)     _mm512_cvtph_ps(_mm256_set_m128i(_mm_loadu_si128((const __m128i *)(y)), _mm_loadu_si128((const __m128i *)(x))))
+#define GGML_F32Cx16_REPEAT_LOAD(x)  _mm512_cvtph_ps(_mm256_set_m128i(x, x))
+#endif
 // the  _mm256_cvt intrinsics require F16C
 #define GGML_F32Cx8_LOAD(x)     _mm256_cvtph_ps(_mm_loadu_si128((const __m128i *)(x)))
 #define GGML_F32Cx8_REPEAT_LOAD(x, loadMask)     _mm256_cvtph_ps(_mm_shuffle_epi32(_mm_maskload_epi32((int const*)(x), loadMask), 68))
 #define GGML_F32Cx8_REARRANGE_LOAD(x, arrangeMask)     _mm256_cvtph_ps(_mm_shuffle_epi8(_mm_loadu_si128((const __m128i *) x), arrangeMask))
 #else
+#if defined(__AVX512F__)
+static inline __m512 __avx512_f32cx8x2_load(ggml_fp16_t *x, ggml_fp16_t *y) {
+    float tmp[16];
+
+    for (int i = 0; i < 8; i++) {
+        tmp[i] = GGML_FP16_TO_FP32(x[i]);
+    }
+
+    for (int i = 0; i < 8; i++) {
+        tmp[i + 8] = GGML_FP16_TO_FP32(y[i]);
+    }
+
+    return _mm512_loadu_ps(tmp);
+}
+static inline __m512 __avx512_repeat_f32cx16_load(__m128i x) {
+    float tmp[16];
+    uint16_t tmphalf[8];
+    _mm_storeu_si128((__m128i*)tmphalf, x);
+
+    for (int i = 0; i < 4; i++) {
+        tmp[i] = GGML_FP16_TO_FP32(tmphalf[i]);
+        tmp[i + 4] = GGML_FP16_TO_FP32(tmphalf[i]);
+        tmp[i + 8] = GGML_FP16_TO_FP32(tmphalf[i]);
+        tmp[i + 12] = GGML_FP16_TO_FP32(tmphalf[i]);
+    }
+
+    return _mm512_loadu_ps(tmp);
+}
+#endif
 static inline __m256 __avx_f32cx8_load(ggml_fp16_t *x) {
     float tmp[8];
 
@@ -78,30 +114,65 @@ static inline __m256 __avx_rearranged_f32cx8_load(ggml_fp16_t *x, __m128i arrang
 #define GGML_F32Cx8_LOAD(x)     __avx_f32cx8_load(x)
 #define GGML_F32Cx8_REPEAT_LOAD(x, loadMask)     __avx_repeat_f32cx8_load(x)
 #define GGML_F32Cx8_REARRANGE_LOAD(x, arrangeMask)     __avx_rearranged_f32cx8_load(x, arrangeMask)
+#if defined(__AVX512F__)
+#define GGML_F32Cx8x2_LOAD(x, y)     __avx512_f32cx8x2_load(x, y)
+#define GGML_F32Cx16_REPEAT_LOAD(x)  __avx512_repeat_f32cx16_load(x)
+#endif
 #endif
 #endif
 
 
 #if defined(__AVX2__) || defined(__AVX512F__)
-static inline __m256i sum_i16_pairs_int(const __m256i x) {
+#if defined(__AVX512F__)
+// add int16_t pairwise and return as 512 bit int vector
+static inline __m512i sum_i16_pairs_int_32x16(const __m512i x) {
+    const __m512i ones = _mm512_set1_epi16(1);
+    return _mm512_madd_epi16(ones, x);
+}
+
+static inline __m512i mul_sum_us8_pairs_int32x16(const __m512i ax, const __m512i sy) {
+#if defined(__AVXVNNI__) || (defined(__AVX512VNNI__) && defined(__AVX512VL__))
+    const __m512i zero = _mm512_setzero_si512();
+    return _mm512_dpbusd_epi32(zero, ax, sy);
+#else
+    // Perform multiplication and create 16-bit values
+    const __m512i dot = _mm512_maddubs_epi16(ax, sy);
+    return sum_i16_pairs_int_32x16(dot);
+#endif
+}
+
+// multiply int8_t, add results pairwise twice and return as 512 bit int vector
+static inline __m512i mul_sum_i8_pairs_int32x16(const __m512i x, const __m512i y) {
+    const __m512i zero = _mm512_setzero_si512();
+    // Get absolute values of x vectors
+    const __m512i ax = _mm512_abs_epi8(x);
+    // Sign the values of the y vectors
+    __mmask64 blt0 = _mm512_movepi8_mask(x);
+    const __m512i sy = _mm512_mask_sub_epi8(y, blt0, zero, y);
+    return mul_sum_us8_pairs_int32x16(ax, sy);
+}
+#endif
+
+// add int16_t pairwise and return as 256 bit int vector
+static inline __m256i sum_i16_pairs_int32x8(const __m256i x) {
     const __m256i ones = _mm256_set1_epi16(1);
     return _mm256_madd_epi16(ones, x);
 }
 
-static inline __m256i mul_sum_us8_pairs_int(const __m256i ax, const __m256i sy) {
+static inline __m256i mul_sum_us8_pairs_int32x8(const __m256i ax, const __m256i sy) {
 #if defined(__AVXVNNI__) || (defined(__AVX512VNNI__) && defined(__AVX512VL__))
     const __m256i zero = _mm256_setzero_si256();
     return _mm256_dpbusd_epi32(zero, ax, sy);
 #else
     // Perform multiplication and create 16-bit values
     const __m256i dot = _mm256_maddubs_epi16(ax, sy);
-    return sum_i16_pairs_int(dot);
+    return sum_i16_pairs_int32x8(dot);
 #endif
 }
 
 // Integer variant of the function defined in ggml-quants.c
-// multiply int8_t, add results pairwise twice and return as float vector
+// multiply int8_t, add results pairwise twice and return as 256 bit int vector
-static inline __m256i mul_sum_i8_pairs_int(const __m256i x, const __m256i y) {
+static inline __m256i mul_sum_i8_pairs_int32x8(const __m256i x, const __m256i y) {
 #if __AVXVNNIINT8__
     const __m256i zero = _mm256_setzero_si256();
     return _mm256_dpbssd_epi32(zero, x, y);
@@ -110,7 +181,7 @@ static inline __m256i mul_sum_i8_pairs_int(const __m256i x, const __m256i y) {
     const __m256i ax = _mm256_sign_epi8(x, x);
     // Sign the values of the y vectors
     const __m256i sy = _mm256_sign_epi8(y, x);
-    return mul_sum_us8_pairs_int(ax, sy);
+    return mul_sum_us8_pairs_int32x8(ax, sy);
 #endif
 }
 #endif
@@ -546,16 +617,8 @@ void ggml_gemv_q4_0_4x4_q8_0(int n, float * restrict s, size_t bs, const void *
     UNUSED(ncols_interleaved);
     UNUSED(blocklen);
 
-#if defined(__ARM_FEATURE_SVE)
+#if ! ((defined(_MSC_VER)) && ! defined(__clang__)) && defined(__aarch64__) && defined(__ARM_NEON)
-    if (ggml_sve_cnt_b == QK8_0) {
+    if (ggml_cpu_has_neon()) {
-        GGML_ASSERT(!(ggml_cpu_has_sve() && (ggml_sve_cnt_b == QK8_0)) &&
-                    "__ARM_FEATURE_SVE defined, use the Q4_0_8_8 quantization format for optimal performance");
-    }
-#endif
-#if defined(__ARM_NEON) && defined(__ARM_FEATURE_MATMUL_INT8)
-    GGML_ASSERT(!(ggml_cpu_has_neon() && ggml_cpu_has_matmul_int8()) &&
-                "__ARM_NEON and __ARM_FEATURE_MATMUL_INT8 defined, use the Q4_0_4_8 quantization format for optimal performance");
-#elif defined(__ARM_NEON) && defined(__aarch64__) && ! ((defined(_MSC_VER)) && ! defined(__clang__))
         const void * b_ptr = vx;
         const void * a_ptr = vy;
         float * res_ptr = s;
@@ -612,7 +675,9 @@ void ggml_gemv_q4_0_4x4_q8_0(int n, float * restrict s, size_t bs, const void *
             : [a_ptr] "r" (a_ptr), [nb] "r" (nb)
             : "memory", "v16", "v17", "v18", "v19", "v20", "v21", "v22", "v23", "v24", "v25", "v26", "v27", "v28", "v29", "v30", "v31", "x20", "x21", "x22"
             );
-#else
+        return;
+    }
+#endif // #if ! ((defined(_MSC_VER)) && ! defined(__clang__)) && defined(__aarch64__) && defined(__ARM_NEON)
     float sumf[4];
     int sumi;
 
@@ -636,7 +701,6 @@ void ggml_gemv_q4_0_4x4_q8_0(int n, float * restrict s, size_t bs, const void *
         }
         for (int j = 0; j < ncols_interleaved; j++) s[x * ncols_interleaved + j] = sumf[j];
     }
-#endif
 }
 
 void ggml_gemv_q4_0_4x8_q8_0(int n, float * restrict s, size_t bs, const void * restrict vx, const void * restrict vy, int nr, int nc) {
@@ -658,13 +722,8 @@ void ggml_gemv_q4_0_4x8_q8_0(int n, float * restrict s, size_t bs, const void *
     UNUSED(ncols_interleaved);
     UNUSED(blocklen);
 
-#if defined(__ARM_FEATURE_SVE)
+#if ! ((defined(_MSC_VER)) && ! defined(__clang__)) && defined(__aarch64__) && defined(__ARM_NEON) && defined(__ARM_FEATURE_MATMUL_INT8)
-    if (ggml_sve_cnt_b == QK8_0) {
+    if (ggml_cpu_has_neon() && ggml_cpu_has_matmul_int8()) {
-        GGML_ASSERT(!(ggml_cpu_has_sve() && (ggml_sve_cnt_b == QK8_0)) &&
-                    "__ARM_FEATURE_SVE defined, use the Q4_0_8_8 quantization format for optimal performance");
-    }
-#endif
-#if defined(__ARM_NEON) && defined(__ARM_FEATURE_MATMUL_INT8) && ! ((defined(_MSC_VER)) && ! defined(__clang__))
         const void * b_ptr = vx;
         const void * a_ptr = vy;
         float * res_ptr = s;
@@ -726,11 +785,9 @@ void ggml_gemv_q4_0_4x8_q8_0(int n, float * restrict s, size_t bs, const void *
             : [a_ptr] "r" (a_ptr), [nb] "r" (nb)
             : "memory", "v0", "v1", "v2", "v16", "v17", "v18", "v19", "v20", "v21", "v22", "v23", "v24", "v25", "v26", "v27", "v28", "v29", "v30", "v31", "x20", "x21", "x22", "x23"
         );
-#elif defined(__ARM_NEON) && defined(__aarch64__)
+        return;
-    GGML_ASSERT((ggml_cpu_has_sve() || ggml_cpu_has_matmul_int8()) &&
+    }
-                "__ARM_FEATURE_SVE and __ARM_FEATURE_MATMUL_INT8 not defined, use the Q4_0_4_4 quantization format for optimal "
+#endif // #if ! ((defined(_MSC_VER)) && ! defined(__clang__)) && defined(__aarch64__) && defined(__ARM_NEON) && defined(__ARM_FEATURE_MATMUL_INT8)
-                "performance");
-#else
     float sumf[4];
     int sumi;
 
@@ -754,7 +811,6 @@ void ggml_gemv_q4_0_4x8_q8_0(int n, float * restrict s, size_t bs, const void *
         }
         for (int j = 0; j < ncols_interleaved; j++) s[x * ncols_interleaved + j] = sumf[j];
     }
-#endif
 }
 
 void ggml_gemv_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void * restrict vx, const void * restrict vy, int nr, int nc) {
@@ -776,8 +832,9 @@ void ggml_gemv_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
     UNUSED(ncols_interleaved);
     UNUSED(blocklen);
 
-#if defined(__ARM_FEATURE_SVE) && ! ((defined(_MSC_VER)) && ! defined(__clang__))
+#if ! ((defined(_MSC_VER)) && ! defined(__clang__)) && defined(__aarch64__)
-    if (ggml_sve_cnt_b == QK8_0) {
+#if defined(__ARM_FEATURE_SVE)
+    if (ggml_cpu_has_sve() && ggml_cpu_get_sve_cnt() == QK8_0) {
         const void * b_ptr = vx;
         const void * a_ptr = vy;
         float * res_ptr = s;
@@ -842,24 +899,7 @@ void ggml_gemv_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
         );
         return;
     }
-    else if (ggml_cpu_has_neon() && ggml_cpu_has_matmul_int8()) {
+#endif // #if defined(__ARM_FEATURE_SVE)
-        GGML_ASSERT((ggml_cpu_has_sve() && (ggml_sve_cnt_b == QK8_0)) &&
-                    "__ARM_FEATURE_SVE for vector size of 256-bits not defined, use the Q4_0_4_8 quantization format for optimal "
-                    "performance");
-    }
-    else if (ggml_cpu_has_neon()) {
-        GGML_ASSERT(((ggml_cpu_has_sve() && (ggml_sve_cnt_b == QK8_0)) || ggml_cpu_has_matmul_int8()) &&
-                    "__ARM_FEATURE_SVE for vector size of 256-bits and __ARM_FEATURE_MATMUL_INT8 not defined, use the Q4_0_4_4 "
-                    "quantization format for optimal performance");
-    }
-#endif
-#if defined(__ARM_NEON) && defined(__ARM_FEATURE_MATMUL_INT8)
-    GGML_ASSERT(ggml_cpu_has_sve() &&
-                "__ARM_FEATURE_SVE not defined, use the Q4_0_4_8 quantization format for optimal performance");
-#elif defined(__ARM_NEON) && defined(__aarch64__)
-    GGML_ASSERT((ggml_cpu_has_sve() || ggml_cpu_has_matmul_int8()) &&
-                "__ARM_FEATURE_SVE and __ARM_FEATURE_MATMUL_INT8 not defined, use the Q4_0_4_4 quantization format for optimal "
-                "performance");
 #elif defined(__AVX2__)
     // Lookup table to convert signed nibbles to signed bytes
     __m256i signextendlut = _mm256_castsi128_si256(_mm_set_epi8(-1, -2, -3, -4, -5, -6, -7, -8, 7, 6, 5, 4, 3, 2, 1, 0));
@@ -929,17 +969,17 @@ void ggml_gemv_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
                 // ...........................................................................
                 // B0(28-31) B4(28-31) B1(28-31) B5(28-31) B2(28-31) B6(28-31) B3(28-31) B7(28-31) with A0(28-31)
 
-                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int(_mm256_blend_epi32(rhs_vec_0123_0 ,_mm256_shuffle_epi32(rhs_vec_4567_0, 177), 170), _mm256_shuffle_epi32(lhs_vec_0, 0)));
+                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int32x8(_mm256_blend_epi32(rhs_vec_0123_0 ,_mm256_shuffle_epi32(rhs_vec_4567_0, 177), 170), _mm256_shuffle_epi32(lhs_vec_0, 0)));
-                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int(_mm256_blend_epi32(_mm256_shuffle_epi32(rhs_vec_0123_0, 177) ,rhs_vec_4567_0, 170), _mm256_shuffle_epi32(lhs_vec_0, 85)));
+                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int32x8(_mm256_blend_epi32(_mm256_shuffle_epi32(rhs_vec_0123_0, 177) ,rhs_vec_4567_0, 170), _mm256_shuffle_epi32(lhs_vec_0, 85)));
 
-                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int(_mm256_blend_epi32(rhs_vec_0123_1 ,_mm256_shuffle_epi32(rhs_vec_4567_1, 177), 170), _mm256_shuffle_epi32(lhs_vec_0, 170)));
+                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int32x8(_mm256_blend_epi32(rhs_vec_0123_1 ,_mm256_shuffle_epi32(rhs_vec_4567_1, 177), 170), _mm256_shuffle_epi32(lhs_vec_0, 170)));
-                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int(_mm256_blend_epi32(_mm256_shuffle_epi32(rhs_vec_0123_1, 177) ,rhs_vec_4567_1, 170), _mm256_shuffle_epi32(lhs_vec_0, 255)));
+                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int32x8(_mm256_blend_epi32(_mm256_shuffle_epi32(rhs_vec_0123_1, 177) ,rhs_vec_4567_1, 170), _mm256_shuffle_epi32(lhs_vec_0, 255)));
 
-                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int(_mm256_blend_epi32(rhs_vec_0123_2 ,_mm256_shuffle_epi32(rhs_vec_4567_2, 177), 170), _mm256_shuffle_epi32(lhs_vec_1, 0)));
+                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int32x8(_mm256_blend_epi32(rhs_vec_0123_2 ,_mm256_shuffle_epi32(rhs_vec_4567_2, 177), 170), _mm256_shuffle_epi32(lhs_vec_1, 0)));
-                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int(_mm256_blend_epi32(_mm256_shuffle_epi32(rhs_vec_0123_2, 177) ,rhs_vec_4567_2, 170), _mm256_shuffle_epi32(lhs_vec_1, 85)));
+                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int32x8(_mm256_blend_epi32(_mm256_shuffle_epi32(rhs_vec_0123_2, 177) ,rhs_vec_4567_2, 170), _mm256_shuffle_epi32(lhs_vec_1, 85)));
 
-                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int(_mm256_blend_epi32(rhs_vec_0123_3 ,_mm256_shuffle_epi32(rhs_vec_4567_3, 177), 170), _mm256_shuffle_epi32(lhs_vec_1, 170)));
+                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int32x8(_mm256_blend_epi32(rhs_vec_0123_3 ,_mm256_shuffle_epi32(rhs_vec_4567_3, 177), 170), _mm256_shuffle_epi32(lhs_vec_1, 170)));
-                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int(_mm256_blend_epi32(_mm256_shuffle_epi32(rhs_vec_0123_3, 177) ,rhs_vec_4567_3, 170), _mm256_shuffle_epi32(lhs_vec_1, 255)));
+                iacc = _mm256_add_epi32(iacc, mul_sum_i8_pairs_int32x8(_mm256_blend_epi32(_mm256_shuffle_epi32(rhs_vec_0123_3, 177) ,rhs_vec_4567_3, 170), _mm256_shuffle_epi32(lhs_vec_1, 255)));
 
                 // Accumulated values multipled with appropriate scales
                 acc_row = _mm256_fmadd_ps(_mm256_cvtepi32_ps(iacc), _mm256_mul_ps(col_scale_f32, row_scale_f32), acc_row);
@@ -950,7 +990,9 @@ void ggml_gemv_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
             _mm256_storeu_ps(s + (y * nr + x * 8), acc_row);
         }
     }
-#else
+    return;
+#endif // #if ! ((defined(_MSC_VER)) && ! defined(__clang__)) && defined(__aarch64__)
+    {
         float sumf[8];
         int sumi;
 
@@ -974,7 +1016,7 @@ void ggml_gemv_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
             }
             for (int j = 0; j < ncols_interleaved; j++) s[x * ncols_interleaved + j] = sumf[j];
         }
-#endif
+    }
 }
 
 void ggml_gemm_q4_0_4x4_q8_0(int n, float * restrict s, size_t bs, const void * restrict vx, const void * restrict vy, int nr, int nc) {
@@ -997,16 +1039,8 @@ void ggml_gemm_q4_0_4x4_q8_0(int n, float * restrict s, size_t bs, const void *
     UNUSED(ncols_interleaved);
     UNUSED(blocklen);
 
-#if defined(__ARM_FEATURE_SVE) && defined(__ARM_FEATURE_MATMUL_INT8)
+#if ! ((defined(_MSC_VER)) && ! defined(__clang__)) && defined(__aarch64__) && defined(__ARM_NEON)
-    if (ggml_sve_cnt_b == QK8_0) {
+    if (ggml_cpu_has_neon()) {
-        GGML_ASSERT(!(ggml_cpu_has_sve() && (ggml_sve_cnt_b == QK8_0)) &&
-                    "__ARM_FEATURE_SVE defined, use the Q4_0_8_8 quantization format for optimal performance");
-    }
-#endif
-#if defined(__ARM_NEON) && defined(__ARM_FEATURE_MATMUL_INT8)
-    GGML_ASSERT(!(ggml_cpu_has_neon() && ggml_cpu_has_matmul_int8()) &&
-                "__ARM_NEON and __ARM_FEATURE_MATMUL_INT8 defined, use the Q4_0_4_8 quantization format for optimal performance");
-#elif defined(__ARM_NEON) && defined(__aarch64__) && ! ((defined(_MSC_VER)) && ! defined(__clang__))
         const void * b_ptr = vx;
         const void * a_ptr = vy;
         float * res_ptr = s;
@@ -1462,7 +1496,10 @@ void ggml_gemm_q4_0_4x4_q8_0(int n, float * restrict s, size_t bs, const void *
             : [b_ptr] "r" (b_ptr), [nr] "r" (nr), [nb] "r" (nb), [res_stride] "r" (res_stride), [nc] "r" (nc)
             : "cc", "memory", "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v8", "v9", "v10", "v11", "v12", "v13", "v14", "v15", "v16", "v17", "v18", "v19", "v20", "v21", "v22", "v23", "v24", "v25", "v26", "v27", "v28", "v29", "v30", "v31", "x9", "x10", "x20", "x21", "x22", "x23", "x24", "x25", "x26", "x27", "x28"
         );
-#else
+        return;
+    }
+#endif // #if ! ((defined(_MSC_VER)) && ! defined(__clang__)) && defined(__aarch64__) && defined(__ARM_NEON)
+    {
         float sumf[4][4];
         int sumi;
 
@@ -1495,7 +1532,7 @@ void ggml_gemm_q4_0_4x4_q8_0(int n, float * restrict s, size_t bs, const void *
                 }
             }
         }
-#endif
+    }
 }
 
 void ggml_gemm_q4_0_4x8_q8_0(int n, float * restrict s, size_t bs, const void * restrict vx, const void * restrict vy, int nr, int nc) {
@@ -1518,13 +1555,8 @@ void ggml_gemm_q4_0_4x8_q8_0(int n, float * restrict s, size_t bs, const void *
     UNUSED(ncols_interleaved);
     UNUSED(blocklen);
 
-#if defined(__ARM_FEATURE_SVE) && defined(__ARM_FEATURE_MATMUL_INT8)
+#if ! ((defined(_MSC_VER)) && ! defined(__clang__)) && defined(__aarch64__) && defined(__ARM_NEON) && defined(__ARM_FEATURE_MATMUL_INT8)
-    if (ggml_sve_cnt_b == QK8_0) {
+    if (ggml_cpu_has_neon() && ggml_cpu_has_matmul_int8()) {
-        GGML_ASSERT(!(ggml_cpu_has_sve() && (ggml_sve_cnt_b == QK8_0)) &&
-                    "__ARM_FEATURE_SVE defined, use the Q4_0_8_8 quantization format for optimal performance");
-    }
-#endif
-#if defined(__ARM_NEON) && defined(__ARM_FEATURE_MATMUL_INT8) && ! ((defined(_MSC_VER)) && ! defined(__clang__))
         const void * b_ptr = vx;
         const void * a_ptr = vy;
         float * res_ptr = s;
@@ -1920,11 +1952,9 @@ void ggml_gemm_q4_0_4x8_q8_0(int n, float * restrict s, size_t bs, const void *
             : [b_ptr] "r" (b_ptr), [nr] "r" (nr), [nb] "r" (nb), [res_stride] "r" (res_stride), [nc] "r" (nc)
             : "cc", "memory", "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v8", "v9", "v10", "v11", "v12", "v13", "v14", "v15", "v16", "v17", "v18", "v19", "v20", "v21", "v22", "v23", "v24", "v25", "v26", "v27", "v28", "v29", "v30", "v31", "x9", "x10", "x20", "x21", "x22", "x23", "x24", "x25", "x26", "x27", "x28"
         );
-#elif defined(__ARM_NEON) && defined(__aarch64__)
+        return;
-    GGML_ASSERT((ggml_cpu_has_sve() || ggml_cpu_has_matmul_int8()) &&
+    }
-                "__ARM_FEATURE_SVE and __ARM_FEATURE_MATMUL_INT8 not defined, use the Q4_0_4_4 quantization format for optimal "
+#endif // #if ! ((defined(_MSC_VER)) && ! defined(__clang__)) && defined(__aarch64__) && defined(__ARM_NEON) && defined(__ARM_FEATURE_MATMUL_INT8)
-                "performance");
-#else
     float sumf[4][4];
     int sumi;
 
@@ -1957,7 +1987,6 @@ void ggml_gemm_q4_0_4x8_q8_0(int n, float * restrict s, size_t bs, const void *
             }
         }
     }
-#endif
 }
 
 void ggml_gemm_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void * restrict vx, const void * restrict vy, int nr, int nc) {
@@ -1980,8 +2009,9 @@ void ggml_gemm_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
     UNUSED(ncols_interleaved);
     UNUSED(blocklen);
 
-#if defined(__ARM_FEATURE_SVE) && defined(__ARM_FEATURE_MATMUL_INT8) && ! ((defined(_MSC_VER)) && ! defined(__clang__))
+#if ! ((defined(_MSC_VER)) && ! defined(__clang__)) && defined(__aarch64__)
-    if (ggml_sve_cnt_b == QK8_0) {
+#if defined(__ARM_FEATURE_SVE) && defined(__ARM_FEATURE_MATMUL_INT8)
+    if (ggml_cpu_has_sve() && ggml_cpu_has_matmul_int8() && ggml_cpu_get_sve_cnt() == QK8_0) {
         const void * b_ptr = vx;
         const void * a_ptr = vy;
         float * res_ptr = s;
@@ -2391,25 +2421,9 @@ void ggml_gemm_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
         );
         return;
     }
-    else if (ggml_cpu_has_neon() && ggml_cpu_has_matmul_int8()) {
+#endif // #if defined(__ARM_FEATURE_SVE) && defined(__ARM_FEATURE_MATMUL_INT8)
-        GGML_ASSERT((ggml_cpu_has_sve() && (ggml_sve_cnt_b == QK8_0)) &&
-                    "__ARM_FEATURE_SVE for vector size of 256-bits not defined, use the Q4_0_4_8 quantization format for optimal "
-                    "performance");
-    }
-    else if (ggml_cpu_has_neon()) {
-        GGML_ASSERT(((ggml_cpu_has_sve() && (ggml_sve_cnt_b == QK8_0)) || ggml_cpu_has_matmul_int8()) &&
-                    "__ARM_FEATURE_SVE for vector size of 256-bits and __ARM_FEATURE_MATMUL_INT8 not defined, use the Q4_0_4_4 "
-                    "quantization format for optimal performance");
-    }
-#endif
-#if defined(__ARM_NEON) && defined(__ARM_FEATURE_MATMUL_INT8)
-    GGML_ASSERT(ggml_cpu_has_sve() &&
-                "__ARM_FEATURE_SVE not defined, use the Q4_0_4_8 quantization format for optimal performance");
-#elif defined(__ARM_NEON) && defined(__aarch64__)
-    GGML_ASSERT((ggml_cpu_has_sve() || ggml_cpu_has_matmul_int8()) &&
-                "__ARM_FEATURE_SVE and __ARM_FEATURE_MATMUL_INT8 not defined, use the Q4_0_4_4 quantization format for optimal "
-                "performance");
 #elif defined(__AVX2__) || defined(__AVX512F__)
+    {
         const block_q4_0x8 * b_ptr_start = (const block_q4_0x8 *)vx;
         const block_q8_0x4 * a_ptr_start = (const block_q8_0x4 *)vy;
         int64_t b_nb = n / QK4_0;
@@ -2421,10 +2435,411 @@ void ggml_gemm_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
         __m256i signextendlut = _mm256_castsi128_si256(_mm_set_epi8(-1, -2, -3, -4, -5, -6, -7, -8, 7, 6, 5, 4, 3, 2, 1, 0));
         signextendlut = _mm256_permute2f128_si256(signextendlut, signextendlut, 0);
         // Permute mask used for easier vector processing at later stages
-    __m256i requiredOrder = _mm256_set_epi32(3 ,2 ,1 ,0, 7 ,6, 5, 4);
+        __m256i requiredOrder = _mm256_set_epi32(3, 2, 1, 0, 7, 6, 5, 4);
+        int64_t xstart = 0;
+        int anr = nr - nr%16; // Used to align nr with boundary of 16
+    #ifdef __AVX512F__
+        int anc = nc - nc%16; // Used to align nc with boundary of 16
+        // Mask to mask out nibbles from packed bytes expanded to 512 bit length
+        const __m512i m4bexpanded = _mm512_set1_epi8(0x0F);
+        // Lookup table to convert signed nibbles to signed bytes expanded to 512 bit length
+        __m512i signextendlutexpanded = _mm512_inserti32x8(_mm512_castsi256_si512(signextendlut), signextendlut, 1);
+
+        // Take group of four block_q8_0x4 structures at each pass of the loop and perform dot product operation
+        for (; y < anr / 4; y += 4) {
+
+            const block_q8_0x4 * a_ptrs[4];
+
+            a_ptrs[0] = a_ptr_start + (y * nb);
+            for (int i = 0; i < 3; ++i) {
+                a_ptrs[i + 1] = a_ptrs[i] + nb;
+            }
+
+            // Take group of two block_q4_0x8 structures at each pass of the loop and perform dot product operation
+            for (int64_t x = 0; x < anc / 8; x += 2) {
+
+                const block_q4_0x8 * b_ptr_0 = b_ptr_start + ((x)     * b_nb);
+                const block_q4_0x8 * b_ptr_1 = b_ptr_start + ((x + 1) * b_nb);
+
+                // Master FP accumulators
+                __m512 acc_rows[16];
+                for (int i = 0; i < 16; i++) {
+                    acc_rows[i] = _mm512_setzero_ps();
+                }
+
+                for (int64_t b = 0; b < nb; b++) {
+                    // Load the sixteen block_q4_0 quantized values interleaved with each other in chunks of eight - B0,B1 ....BE,BF
+                    const __m256i rhs_raw_mat_0123_0 = _mm256_loadu_si256((const __m256i *)(b_ptr_0[b].qs));
+                    const __m256i rhs_raw_mat_4567_0 = _mm256_loadu_si256((const __m256i *)(b_ptr_0[b].qs + 32));
+                    const __m256i rhs_raw_mat_0123_1 = _mm256_loadu_si256((const __m256i *)(b_ptr_0[b].qs + 64));
+                    const __m256i rhs_raw_mat_4567_1 = _mm256_loadu_si256((const __m256i *)(b_ptr_0[b].qs + 96));
+
+                    const __m256i rhs_raw_mat_89AB_0 = _mm256_loadu_si256((const __m256i *)(b_ptr_1[b].qs));
+                    const __m256i rhs_raw_mat_CDEF_0 = _mm256_loadu_si256((const __m256i *)(b_ptr_1[b].qs + 32));
+                    const __m256i rhs_raw_mat_89AB_1 = _mm256_loadu_si256((const __m256i *)(b_ptr_1[b].qs + 64));
+                    const __m256i rhs_raw_mat_CDEF_1 = _mm256_loadu_si256((const __m256i *)(b_ptr_1[b].qs + 96));
+
+                    // Save the values in the following vectors in the formats B0B1B4B5B8B9BCBD, B2B3B6B7BABBBEBF for further processing and storing of values
+                    const __m256i rhs_raw_mat_0145_0 = _mm256_blend_epi32(rhs_raw_mat_0123_0, _mm256_permutevar8x32_epi32(rhs_raw_mat_4567_0, requiredOrder), 240);
+                    const __m256i rhs_raw_mat_2367_0 = _mm256_blend_epi32(_mm256_permutevar8x32_epi32(rhs_raw_mat_0123_0, requiredOrder), rhs_raw_mat_4567_0, 240);
+                    const __m256i rhs_raw_mat_0145_1 = _mm256_blend_epi32(rhs_raw_mat_0123_1, _mm256_permutevar8x32_epi32(rhs_raw_mat_4567_1, requiredOrder), 240);
+                    const __m256i rhs_raw_mat_2367_1 = _mm256_blend_epi32(_mm256_permutevar8x32_epi32(rhs_raw_mat_0123_1, requiredOrder), rhs_raw_mat_4567_1, 240);
+
+                    const __m256i rhs_raw_mat_89CD_0 = _mm256_blend_epi32(rhs_raw_mat_89AB_0, _mm256_permutevar8x32_epi32(rhs_raw_mat_CDEF_0, requiredOrder), 240);
+                    const __m256i rhs_raw_mat_ABEF_0 = _mm256_blend_epi32(_mm256_permutevar8x32_epi32(rhs_raw_mat_89AB_0, requiredOrder), rhs_raw_mat_CDEF_0, 240);
+                    const __m256i rhs_raw_mat_89CD_1 = _mm256_blend_epi32(rhs_raw_mat_89AB_1, _mm256_permutevar8x32_epi32(rhs_raw_mat_CDEF_1, requiredOrder), 240);
+                    const __m256i rhs_raw_mat_ABEF_1 = _mm256_blend_epi32(_mm256_permutevar8x32_epi32(rhs_raw_mat_89AB_1, requiredOrder), rhs_raw_mat_CDEF_1, 240);
+
+                    const __m512i rhs_raw_mat_014589CD_0 = _mm512_inserti32x8(_mm512_castsi256_si512(rhs_raw_mat_0145_0), rhs_raw_mat_89CD_0, 1);
+                    const __m512i rhs_raw_mat_2367ABEF_0 = _mm512_inserti32x8(_mm512_castsi256_si512(rhs_raw_mat_2367_0), rhs_raw_mat_ABEF_0, 1);
+                    const __m512i rhs_raw_mat_014589CD_1 = _mm512_inserti32x8(_mm512_castsi256_si512(rhs_raw_mat_0145_1), rhs_raw_mat_89CD_1, 1);
+                    const __m512i rhs_raw_mat_2367ABEF_1 = _mm512_inserti32x8(_mm512_castsi256_si512(rhs_raw_mat_2367_1), rhs_raw_mat_ABEF_1, 1);
+
+                    // 4-bit -> 8-bit - Sign is maintained
+                    const __m512i rhs_mat_014589CD_0 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(rhs_raw_mat_014589CD_0, m4bexpanded)); //B0(0-7) B1(0-7) B4(0-7) B5(0-7) B8(0-7) B9(0-7) BC(0-7) BD(0-7)
+                    const __m512i rhs_mat_2367ABEF_0 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(rhs_raw_mat_2367ABEF_0, m4bexpanded)); //B2(0-7) B3(0-7) B6(0-7) B7(0-7) BA(0-7) BB(0-7) BE(0-7) BF(0-7)
+
+                    const __m512i rhs_mat_014589CD_1 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(rhs_raw_mat_014589CD_1, m4bexpanded)); //B0(8-15) B1(8-15) B4(8-15) B5(8-15) B8(8-15) B9(8-15) BC(8-15) BD(8-15)
+                    const __m512i rhs_mat_2367ABEF_1 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(rhs_raw_mat_2367ABEF_1, m4bexpanded)); //B2(8-15) B3(8-15) B6(8-15) B7(8-15) BA(8-15) BB(8-15) BE(8-15) BF(8-15)
+
+                    const __m512i rhs_mat_014589CD_2 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(_mm512_srli_epi16(rhs_raw_mat_014589CD_0, 4), m4bexpanded)); //B0(16-23) B1(16-23) B4(16-23) B5(16-23) B8(16-23) B9(16-23) BC(16-23) BD(16-23)
+                    const __m512i rhs_mat_2367ABEF_2 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(_mm512_srli_epi16(rhs_raw_mat_2367ABEF_0, 4), m4bexpanded)); //B2(16-23) B3(16-23) B6(16-23) B7(16-23) BA(16-23) BB(16-23) BE(16-23) BF(16-23)
+
+                    const __m512i rhs_mat_014589CD_3 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(_mm512_srli_epi16(rhs_raw_mat_014589CD_1, 4), m4bexpanded)); //B0(24-31) B1(24-31) B4(24-31) B5(24-31) B8(24-31) B9(24-31) BC(24-31) BD(24-31)
+                    const __m512i rhs_mat_2367ABEF_3 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(_mm512_srli_epi16(rhs_raw_mat_2367ABEF_1, 4), m4bexpanded)); //B2(24-31) B3(24-31) B6(24-31) B7(24-31) BA(24-31) BB(24-31) BE(24-31) BF(24-31)
+
+                    // Shuffle pattern one - right side input
+                    const __m512i rhs_mat_014589CD_0_sp1 = _mm512_shuffle_epi32(rhs_mat_014589CD_0, 136); //B0(0-3) B1(0-3) B0(0-3) B1(0-3) B4(0-3) B5(0-3) B4(0-3) B5(0-3) B8(0-3) B9(0-3) B8(0-3) B9(0-3) BC(0-3) BD(0-3) BC(0-3) BD(0-3)
+                    const __m512i rhs_mat_2367ABEF_0_sp1 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_0, 136); //B2(0-3) B3(0-3) B2(0-3) B3(0-3) B6(0-3) B7(0-3) B6(0-3) B7(0-3) BA(0-3) BB(0-3) BA(0-3) BB(0-3) BE(0-3) BF(0-3) BE(0-3) BF(0-3)
+
+                    const __m512i rhs_mat_014589CD_1_sp1 = _mm512_shuffle_epi32(rhs_mat_014589CD_1, 136); //B0(8-11) B1(8-11) B0(8-11) B1(8-11) B4(8-11) B5(8-11) B4(8-11) B5(8-11) B8(8-11) B9(8-11) B8(8-11) B9(8-11) BC(8-11) BD(8-11) BC(8-11) BD(8-11)
+                    const __m512i rhs_mat_2367ABEF_1_sp1 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_1, 136); //B2(8-11) B3(8-11) B2(8-11) B3(8-11) B6(8-11) B7(8-11) B6(8-11) B7(8-11) BA(8-11) BB(8-11) BA(8-11) BB(8-11) BE(8-11) BF(8-11) BE(8-11) BF(8-11)
+
+                    const __m512i rhs_mat_014589CD_2_sp1 = _mm512_shuffle_epi32(rhs_mat_014589CD_2, 136); //B0(16-19) B1(16-19) B0(16-19) B1(16-19) B4(16-19) B5(16-19) B4(16-19) B5(16-19) B8(16-19) B9(16-19) B8(16-19) B9(16-19) BC(16-19) BD(16-19) BC(16-19) BD(16-19)
+                    const __m512i rhs_mat_2367ABEF_2_sp1 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_2, 136); //B2(16-19) B3(16-19) B2(16-19) B3(16-19) B6(16-19) B7(16-19) B6(16-19) B7(16-19) BA(16-19) BB(16-19) BA(16-19) BB(16-19) BE(16-19) BF(16-19) BE(16-19) BF(16-19)
+
+                    const __m512i rhs_mat_014589CD_3_sp1 = _mm512_shuffle_epi32(rhs_mat_014589CD_3, 136); //B0(24-27) B1(24-27) B0(24-27) B1(24-27) B4(24-27) B5(24-27) B4(24-27) B5(24-27) B8(24-27) B9(24-27) B8(24-27) B9(24-27) BC(24-27) BD(24-27) BC(24-27) BD(24-27)
+                    const __m512i rhs_mat_2367ABEF_3_sp1 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_3, 136); //B2(24-27) B3(24-27) B2(24-27) B3(24-27) B6(24-27) B7(24-27) B6(24-27) B7(24-27) BA(24-27) BB(24-27) BA(24-27) BB(24-27) BE(24-27) BF(24-27) BE(24-27) BF(24-27)
+
+                    // Shuffle pattern two - right side input
+
+                    const __m512i rhs_mat_014589CD_0_sp2 = _mm512_shuffle_epi32(rhs_mat_014589CD_0, 221); //B0(4-7) B1(4-7) B0(4-7) B1(4-7) B4(4-7) B5(4-7) B4(4-7) B5(4-7) B8(4-7) B9(4-7) B8(4-7) B9(4-7) BC(4-7) BD(4-7) BC(4-7) BD(4-7)
+                    const __m512i rhs_mat_2367ABEF_0_sp2 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_0, 221); //B2(4-7) B3(4-7) B2(4-7) B3(4-7) B6(4-7) B7(4-7) B6(4-7) B7(4-7) BA(4-7) BB(4-7) BA(4-7) BB(4-7) BE(4-7) BF(4-7) BE(4-7) BF(4-7)
+
+                    const __m512i rhs_mat_014589CD_1_sp2 = _mm512_shuffle_epi32(rhs_mat_014589CD_1, 221); //B0(12-15) B1(12-15) B0(12-15) B1(12-15) B4(12-15) B5(12-15) B4(12-15) B5(12-15) B8(12-15) B9(12-15) B8(12-15) B9(12-15) BC(12-15) BD(12-15) BC(12-15) BD(12-15)
+                    const __m512i rhs_mat_2367ABEF_1_sp2 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_1, 221); //B2(12-15) B3(12-15) B2(12-15) B3(12-15) B6(12-15) B7(12-15) B6(12-15) B7(12-15) BA(12-15) BB(12-15) BA(12-15) BB(12-15) BE(12-15) BF(12-15) BE(12-15) BF(12-15)
+
+                    const __m512i rhs_mat_014589CD_2_sp2 = _mm512_shuffle_epi32(rhs_mat_014589CD_2, 221); //B0(20-23) B1(20-23) B0(20-23) B1(20-23) B4(20-23) B5(20-23) B4(20-23) B5(20-23) B8(20-23) B9(20-23) B8(20-23) B9(20-23) BC(20-23) BD(20-23) BC(20-23) BD(20-23)
+                    const __m512i rhs_mat_2367ABEF_2_sp2 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_2, 221); //B2(20-23) B3(20-23) B2(20-23) B3(20-23) B6(20-23) B7(20-23) B6(20-23) B7(20-23) BA(20-23) BB(20-23) BA(20-23) BB(20-23) BE(20-23) BF(20-23) BE(20-23) BF(20-23)
+
+                    const __m512i rhs_mat_014589CD_3_sp2 = _mm512_shuffle_epi32(rhs_mat_014589CD_3, 221); //B0(28-31) B1(28-31) B0(28-31) B1(28-31) B4(28-31) B5(28-31) B4(28-31) B5(28-31) B8(28-31) B9(28-31) B8(28-31) B9(28-31) BC(28-31) BD(28-31) BC(28-31) BD(28-31)
+                    const __m512i rhs_mat_2367ABEF_3_sp2 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_3, 221); //B2(28-31) B3(28-31) B2(28-31) B3(28-31) B6(28-31) B7(28-31) B6(28-31) B7(28-31) BA(28-31) BB(28-31) BA(28-31) BB(28-31) BE(28-31) BF(28-31) BE(28-31) BF(28-31)
+
+                    // Scale values - Load the weight scale values of two block_q4_0x8
+                    const __m512 col_scale_f32 = GGML_F32Cx8x2_LOAD(b_ptr_0[b].d, b_ptr_1[b].d);
+
+                    // Process LHS in pairs of rows
+                    for (int rp = 0; rp < 4; rp++) {
+
+                        // Load the four block_q4_0 quantized values interleaved with each other in chunks of eight - A0,A1,A2,A3
+                        // Loaded as set of 128 bit vectors and repeated and stored into a 256 bit vector before again repeating into 512 bit vector
+                        __m256i lhs_mat_ymm_0123_0 = _mm256_loadu_si256((const __m256i *)((a_ptrs[rp][b].qs)));
+                        __m256i lhs_mat_ymm_01_0 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_0, lhs_mat_ymm_0123_0, 0);
+                        __m256i lhs_mat_ymm_23_0 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_0, lhs_mat_ymm_0123_0, 17);
+                        __m256i lhs_mat_ymm_0123_1 = _mm256_loadu_si256((const __m256i *)((a_ptrs[rp][b].qs + 32)));
+                        __m256i lhs_mat_ymm_01_1 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_1, lhs_mat_ymm_0123_1, 0);
+                        __m256i lhs_mat_ymm_23_1 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_1, lhs_mat_ymm_0123_1, 17);
+                        __m256i lhs_mat_ymm_0123_2 = _mm256_loadu_si256((const __m256i *)((a_ptrs[rp][b].qs + 64)));
+                        __m256i lhs_mat_ymm_01_2 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_2, lhs_mat_ymm_0123_2, 0);
+                        __m256i lhs_mat_ymm_23_2 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_2, lhs_mat_ymm_0123_2, 17);
+                        __m256i lhs_mat_ymm_0123_3 = _mm256_loadu_si256((const __m256i *)((a_ptrs[rp][b].qs + 96)));
+                        __m256i lhs_mat_ymm_01_3 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_3, lhs_mat_ymm_0123_3, 0);
+                        __m256i lhs_mat_ymm_23_3 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_3, lhs_mat_ymm_0123_3, 17);
+
+                        __m512i lhs_mat_01_0 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_01_0), lhs_mat_ymm_01_0, 1);
+                        __m512i lhs_mat_23_0 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_23_0), lhs_mat_ymm_23_0, 1);
+                        __m512i lhs_mat_01_1 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_01_1), lhs_mat_ymm_01_1, 1);
+                        __m512i lhs_mat_23_1 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_23_1), lhs_mat_ymm_23_1, 1);
+                        __m512i lhs_mat_01_2 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_01_2), lhs_mat_ymm_01_2, 1);
+                        __m512i lhs_mat_23_2 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_23_2), lhs_mat_ymm_23_2, 1);
+                        __m512i lhs_mat_01_3 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_01_3), lhs_mat_ymm_01_3, 1);
+                        __m512i lhs_mat_23_3 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_23_3), lhs_mat_ymm_23_3, 1);
+
+                        // Shuffle pattern one - left side input
+
+                        const __m512i lhs_mat_01_0_sp1 = _mm512_shuffle_epi32(lhs_mat_01_0, 160);  //A0(0-3) A0(0-3) A1(0-3) A1(0-3) A0(0-3) A0(0-3) A1(0-3) A1(0-3) A0(0-3) A0(0-3) A1(0-3) A1(0-3) A0(0-3) A0(0-3) A1(0-3) A1(0-3)
+                        const __m512i lhs_mat_23_0_sp1 = _mm512_shuffle_epi32(lhs_mat_23_0, 160);  //A2(0-3) A2(0-3) A3(0-3) A3(0-3) A2(0-3) A2(0-3) A3(0-3) A3(0-3) A2(0-3) A2(0-3) A3(0-3) A3(0-3) A2(0-3) A2(0-3) A3(0-3) A3(0-3)
+
+                        const __m512i lhs_mat_01_1_sp1 = _mm512_shuffle_epi32(lhs_mat_01_1, 160);  //A0(8-11) A0(8-11) A1(8-11) A1(8-11) A0(8-11) A0(8-11) A1(8-11) A1(8-11) A0(8-11) A0(8-11) A1(8-11) A1(8-11) A0(8-11) A0(8-11) A1(8-11) A1(8-11)
+                        const __m512i lhs_mat_23_1_sp1 = _mm512_shuffle_epi32(lhs_mat_23_1, 160);  //A2(8-11) A2(8-11) A3(8-11) A3(8-11) A2(8-11) A2(8-11) A3(8-11) A3(8-11) A2(8-11) A2(8-11) A3(8-11) A3(8-11) A2(8-11) A2(8-11) A3(8-11) A3(8-11)
+
+                        const __m512i lhs_mat_01_2_sp1 = _mm512_shuffle_epi32(lhs_mat_01_2, 160);  //A0(16-19) A0(16-19) A1(16-19) A1(16-19) A0(16-19) A0(16-19) A1(16-19) A1(16-19) A0(16-19) A0(16-19) A1(16-19) A1(16-19) A0(16-19) A0(16-19) A1(16-19) A1(16-19)
+                        const __m512i lhs_mat_23_2_sp1 = _mm512_shuffle_epi32(lhs_mat_23_2, 160);  //A2(16-19) A2(16-19) A3(16-19) A3(16-19) A2(16-19) A2(16-19) A3(16-19) A3(16-19) A2(16-19) A2(16-19) A3(16-19) A3(16-19) A2(16-19) A2(16-19) A3(16-19) A3(16-19)
+
+                        const __m512i lhs_mat_01_3_sp1 = _mm512_shuffle_epi32(lhs_mat_01_3, 160);  //A0(24-27) A0(24-27) A1(24-27) A1(24-27) A0(24-27) A0(24-27) A1(24-27) A1(24-27) A0(24-27) A0(24-27) A1(24-27) A1(24-27) A0(24-27) A0(24-27) A1(24-27) A1(24-27)
+                        const __m512i lhs_mat_23_3_sp1 = _mm512_shuffle_epi32(lhs_mat_23_3, 160);  //A2(24-27) A2(24-27) A3(24-27) A3(24-27) A2(24-27) A2(24-27) A3(24-27) A3(24-27) A2(24-27) A2(24-27) A3(24-27) A3(24-27) A2(24-27) A2(24-27) A3(24-27) A3(24-27)
+
+                        // Shuffle pattern two - left side input
+
+                        const __m512i lhs_mat_01_0_sp2 = _mm512_shuffle_epi32(lhs_mat_01_0, 245);  //A0(4-7) A0(4-7) A1(4-7) A1(4-7) A0(4-7) A0(4-7) A1(4-7) A1(4-7) A0(4-7) A0(4-7) A1(4-7) A1(4-7) A0(4-7) A0(4-7) A1(4-7) A1(4-7)
+                        const __m512i lhs_mat_23_0_sp2 = _mm512_shuffle_epi32(lhs_mat_23_0, 245);  //A2(4-7) A2(4-7) A3(4-7) A3(4-7) A2(4-7) A2(4-7) A3(4-7) A3(4-7) A2(4-7) A2(4-7) A3(4-7) A3(4-7) A2(4-7) A2(4-7) A3(4-7) A3(4-7)
+
+                        const __m512i lhs_mat_01_1_sp2 = _mm512_shuffle_epi32(lhs_mat_01_1, 245);  //A0(12-15) A0(12-15) A1(12-15) A1(12-15) A0(12-15) A0(12-15) A1(12-15) A1(12-15) A0(12-15) A0(12-15) A1(12-15) A1(12-15) A0(12-15) A0(12-15) A1(12-15) A1(12-15)
+                        const __m512i lhs_mat_23_1_sp2 = _mm512_shuffle_epi32(lhs_mat_23_1, 245);  //A2(12-15) A2(12-15) A3(12-15) A3(12-15) A2(12-15) A2(12-15) A3(12-15) A3(12-15) A2(12-15) A2(12-15) A3(12-15) A3(12-15) A2(12-15) A2(12-15) A3(12-15) A3(12-15)
+
+                        const __m512i lhs_mat_01_2_sp2 = _mm512_shuffle_epi32(lhs_mat_01_2, 245);  //A0(20-23) A0(20-23) A1(20-23) A1(20-23) A0(20-23) A0(20-23) A1(20-23) A1(20-23) A0(20-23) A0(20-23) A1(20-23) A1(20-23) A0(20-23) A0(20-23) A1(20-23) A1(20-23)
+                        const __m512i lhs_mat_23_2_sp2 = _mm512_shuffle_epi32(lhs_mat_23_2, 245);  //A2(20-23) A2(20-23) A3(20-23) A3(20-23) A2(20-23) A2(20-23) A3(20-23) A3(20-23) A2(20-23) A2(20-23) A3(20-23) A3(20-23) A2(20-23) A2(20-23) A3(20-23) A3(20-23)
+
+                        const __m512i lhs_mat_01_3_sp2 = _mm512_shuffle_epi32(lhs_mat_01_3, 245);  //A0(28-31) A0(28-31) A1(28-31) A1(28-31) A0(28-31) A0(28-31) A1(28-31) A1(28-31) A0(28-31) A0(28-31) A1(28-31) A1(28-31) A0(28-31) A0(28-31) A1(28-31) A1(28-31)
+                        const __m512i lhs_mat_23_3_sp2 = _mm512_shuffle_epi32(lhs_mat_23_3, 245);  //A2(28-31) A2(28-31) A3(28-31) A3(28-31) A2(28-31) A2(28-31) A3(28-31) A3(28-31) A2(28-31) A2(28-31) A3(28-31) A3(28-31) A2(28-31) A2(28-31) A3(28-31) A3(28-31)
+
+                        // The values arranged in shuffle patterns are operated with dot product operation within 32 bit lane i.e corresponding bytes and multiplied and added into 32 bit integers within 32 bit lane
+                        // Resembles MMLAs into 2x2 matrices in ARM Version
+                        __m512i iacc_mat_00_sp1 =
+                            _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_01_3_sp1, rhs_mat_014589CD_3_sp1), mul_sum_i8_pairs_int32x16(lhs_mat_01_2_sp1, rhs_mat_014589CD_2_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_01_1_sp1, rhs_mat_014589CD_1_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_01_0_sp1, rhs_mat_014589CD_0_sp1));
+                        __m512i iacc_mat_01_sp1 =
+                            _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_01_3_sp1, rhs_mat_2367ABEF_3_sp1), mul_sum_i8_pairs_int32x16(lhs_mat_01_2_sp1, rhs_mat_2367ABEF_2_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_01_1_sp1, rhs_mat_2367ABEF_1_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_01_0_sp1, rhs_mat_2367ABEF_0_sp1));
+                        __m512i iacc_mat_10_sp1 =
+                            _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_23_3_sp1, rhs_mat_014589CD_3_sp1), mul_sum_i8_pairs_int32x16(lhs_mat_23_2_sp1, rhs_mat_014589CD_2_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_23_1_sp1, rhs_mat_014589CD_1_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_23_0_sp1, rhs_mat_014589CD_0_sp1));
+                        __m512i iacc_mat_11_sp1 =
+                            _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_23_3_sp1, rhs_mat_2367ABEF_3_sp1), mul_sum_i8_pairs_int32x16(lhs_mat_23_2_sp1, rhs_mat_2367ABEF_2_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_23_1_sp1, rhs_mat_2367ABEF_1_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_23_0_sp1, rhs_mat_2367ABEF_0_sp1));
+                        __m512i iacc_mat_00_sp2 =
+                            _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_01_3_sp2, rhs_mat_014589CD_3_sp2), mul_sum_i8_pairs_int32x16(lhs_mat_01_2_sp2, rhs_mat_014589CD_2_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_01_1_sp2, rhs_mat_014589CD_1_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_01_0_sp2, rhs_mat_014589CD_0_sp2));
+                        __m512i iacc_mat_01_sp2 =
+                            _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_01_3_sp2, rhs_mat_2367ABEF_3_sp2), mul_sum_i8_pairs_int32x16(lhs_mat_01_2_sp2, rhs_mat_2367ABEF_2_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_01_1_sp2, rhs_mat_2367ABEF_1_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_01_0_sp2, rhs_mat_2367ABEF_0_sp2));
+                        __m512i iacc_mat_10_sp2 =
+                            _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_23_3_sp2, rhs_mat_014589CD_3_sp2), mul_sum_i8_pairs_int32x16(lhs_mat_23_2_sp2, rhs_mat_014589CD_2_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_23_1_sp2, rhs_mat_014589CD_1_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_23_0_sp2, rhs_mat_014589CD_0_sp2));
+                        __m512i iacc_mat_11_sp2 =
+                            _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_23_3_sp2, rhs_mat_2367ABEF_3_sp2), mul_sum_i8_pairs_int32x16(lhs_mat_23_2_sp2, rhs_mat_2367ABEF_2_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_23_1_sp2, rhs_mat_2367ABEF_1_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_23_0_sp2, rhs_mat_2367ABEF_0_sp2));
+
+                        // Output of both shuffle patterns are added in order to sum dot product outputs of all 32 values in block
+                        __m512i iacc_mat_00 = _mm512_add_epi32(iacc_mat_00_sp1, iacc_mat_00_sp2);
+                        __m512i iacc_mat_01 = _mm512_add_epi32(iacc_mat_01_sp1, iacc_mat_01_sp2);
+                        __m512i iacc_mat_10 = _mm512_add_epi32(iacc_mat_10_sp1, iacc_mat_10_sp2);
+                        __m512i iacc_mat_11 = _mm512_add_epi32(iacc_mat_11_sp1, iacc_mat_11_sp2);
+
+
+                        // Straighten out to make 4 row vectors
+                        __m512i iacc_row_0 = _mm512_mask_blend_epi32(0xCCCC, iacc_mat_00, _mm512_shuffle_epi32(iacc_mat_01, 78));
+                        __m512i iacc_row_1 = _mm512_mask_blend_epi32(0xCCCC, _mm512_shuffle_epi32(iacc_mat_00, 78), iacc_mat_01);
+                        __m512i iacc_row_2 = _mm512_mask_blend_epi32(0xCCCC, iacc_mat_10, _mm512_shuffle_epi32(iacc_mat_11, 78));
+                        __m512i iacc_row_3 = _mm512_mask_blend_epi32(0xCCCC, _mm512_shuffle_epi32(iacc_mat_10, 78), iacc_mat_11);
+
+                        // Load the scale(d) values for all the 4 Q8_0 blocks and repeat it across lanes
+                        const __m128i row_scale_f16 = _mm_shuffle_epi32(_mm_maskload_epi32((int const*)(a_ptrs[rp][b].d), loadMask), 68);
+                        const __m512 row_scale_f32 = GGML_F32Cx16_REPEAT_LOAD(row_scale_f16);
+
+                        // Multiply with appropiate scales and accumulate
+                        acc_rows[rp * 4]     = _mm512_fmadd_ps(_mm512_cvtepi32_ps(iacc_row_0), _mm512_mul_ps(col_scale_f32, _mm512_shuffle_ps(row_scale_f32, row_scale_f32, 0)),   acc_rows[rp * 4]);
+                        acc_rows[rp * 4 + 1] = _mm512_fmadd_ps(_mm512_cvtepi32_ps(iacc_row_1), _mm512_mul_ps(col_scale_f32, _mm512_shuffle_ps(row_scale_f32, row_scale_f32, 85)),  acc_rows[rp * 4 + 1]);
+                        acc_rows[rp * 4 + 2] = _mm512_fmadd_ps(_mm512_cvtepi32_ps(iacc_row_2), _mm512_mul_ps(col_scale_f32, _mm512_shuffle_ps(row_scale_f32, row_scale_f32, 170)), acc_rows[rp * 4 + 2]);
+                        acc_rows[rp * 4 + 3] = _mm512_fmadd_ps(_mm512_cvtepi32_ps(iacc_row_3), _mm512_mul_ps(col_scale_f32, _mm512_shuffle_ps(row_scale_f32, row_scale_f32, 255)), acc_rows[rp * 4 + 3]);
+                    }
+                }
+
+                // Store the accumulated values
+                for (int i = 0; i < 16; i++) {
+                    _mm512_storeu_ps((float *)(s + ((y * 4 + i) * bs + x * 8)), acc_rows[i]);
+                }
+            }
+        }
+        // Take a block_q8_0x4 structures at each pass of the loop and perform dot product operation
+        for (; y < nr / 4; y ++) {
+
+            const block_q8_0x4 * a_ptr = a_ptr_start + (y * nb);
+
+            // Take group of two block_q4_0x8 structures at each pass of the loop and perform dot product operation
+            for (int64_t x = 0; x < anc / 8; x += 2) {
+
+                const block_q4_0x8 * b_ptr_0 = b_ptr_start + ((x)     * b_nb);
+                const block_q4_0x8 * b_ptr_1 = b_ptr_start + ((x + 1) * b_nb);
+
+                // Master FP accumulators
+                __m512 acc_rows[4];
+                for (int i = 0; i < 4; i++) {
+                    acc_rows[i] = _mm512_setzero_ps();
+                }
+
+                for (int64_t b = 0; b < nb; b++) {
+                    // Load the sixteen block_q4_0 quantized values interleaved with each other in chunks of eight - B0,B1 ....BE,BF
+                    const __m256i rhs_raw_mat_0123_0 = _mm256_loadu_si256((const __m256i *)(b_ptr_0[b].qs));
+                    const __m256i rhs_raw_mat_4567_0 = _mm256_loadu_si256((const __m256i *)(b_ptr_0[b].qs + 32));
+                    const __m256i rhs_raw_mat_0123_1 = _mm256_loadu_si256((const __m256i *)(b_ptr_0[b].qs + 64));
+                    const __m256i rhs_raw_mat_4567_1 = _mm256_loadu_si256((const __m256i *)(b_ptr_0[b].qs + 96));
+
+                    const __m256i rhs_raw_mat_89AB_0 = _mm256_loadu_si256((const __m256i *)(b_ptr_1[b].qs));
+                    const __m256i rhs_raw_mat_CDEF_0 = _mm256_loadu_si256((const __m256i *)(b_ptr_1[b].qs + 32));
+                    const __m256i rhs_raw_mat_89AB_1 = _mm256_loadu_si256((const __m256i *)(b_ptr_1[b].qs + 64));
+                    const __m256i rhs_raw_mat_CDEF_1 = _mm256_loadu_si256((const __m256i *)(b_ptr_1[b].qs + 96));
+
+                    // Save the values in the following vectors in the formats B0B1B4B5, B2B3B6B7 for further processing and storing of valuess
+                    const __m256i rhs_raw_mat_0145_0 = _mm256_blend_epi32(rhs_raw_mat_0123_0, _mm256_permutevar8x32_epi32(rhs_raw_mat_4567_0, requiredOrder), 240);
+                    const __m256i rhs_raw_mat_2367_0 = _mm256_blend_epi32(_mm256_permutevar8x32_epi32(rhs_raw_mat_0123_0, requiredOrder), rhs_raw_mat_4567_0, 240);
+                    const __m256i rhs_raw_mat_0145_1 = _mm256_blend_epi32(rhs_raw_mat_0123_1, _mm256_permutevar8x32_epi32(rhs_raw_mat_4567_1, requiredOrder), 240);
+                    const __m256i rhs_raw_mat_2367_1 = _mm256_blend_epi32(_mm256_permutevar8x32_epi32(rhs_raw_mat_0123_1, requiredOrder), rhs_raw_mat_4567_1, 240);
+
+                    const __m256i rhs_raw_mat_89CD_0 = _mm256_blend_epi32(rhs_raw_mat_89AB_0, _mm256_permutevar8x32_epi32(rhs_raw_mat_CDEF_0, requiredOrder), 240);
+                    const __m256i rhs_raw_mat_ABEF_0 = _mm256_blend_epi32(_mm256_permutevar8x32_epi32(rhs_raw_mat_89AB_0, requiredOrder), rhs_raw_mat_CDEF_0, 240);
+                    const __m256i rhs_raw_mat_89CD_1 = _mm256_blend_epi32(rhs_raw_mat_89AB_1, _mm256_permutevar8x32_epi32(rhs_raw_mat_CDEF_1, requiredOrder), 240);
+                    const __m256i rhs_raw_mat_ABEF_1 = _mm256_blend_epi32(_mm256_permutevar8x32_epi32(rhs_raw_mat_89AB_1, requiredOrder), rhs_raw_mat_CDEF_1, 240);
+
+                    const __m512i rhs_raw_mat_014589CD_0 = _mm512_inserti32x8(_mm512_castsi256_si512(rhs_raw_mat_0145_0), rhs_raw_mat_89CD_0, 1);
+                    const __m512i rhs_raw_mat_2367ABEF_0 = _mm512_inserti32x8(_mm512_castsi256_si512(rhs_raw_mat_2367_0), rhs_raw_mat_ABEF_0, 1);
+                    const __m512i rhs_raw_mat_014589CD_1 = _mm512_inserti32x8(_mm512_castsi256_si512(rhs_raw_mat_0145_1), rhs_raw_mat_89CD_1, 1);
+                    const __m512i rhs_raw_mat_2367ABEF_1 = _mm512_inserti32x8(_mm512_castsi256_si512(rhs_raw_mat_2367_1), rhs_raw_mat_ABEF_1, 1);
+
+                    // 4-bit -> 8-bit - Sign is maintained
+                    const __m512i rhs_mat_014589CD_0 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(rhs_raw_mat_014589CD_0, m4bexpanded)); //B0(0-7) B1(0-7) B4(0-7) B5(0-7) B8(0-7) B9(0-7) BC(0-7) BD(0-7)
+                    const __m512i rhs_mat_2367ABEF_0 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(rhs_raw_mat_2367ABEF_0, m4bexpanded)); //B2(0-7) B3(0-7) B6(0-7) B7(0-7) BA(0-7) BB(0-7) BE(0-7) BF(0-7)
+
+                    const __m512i rhs_mat_014589CD_1 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(rhs_raw_mat_014589CD_1, m4bexpanded)); //B0(8-15) B1(8-15) B4(8-15) B5(8-15) B8(8-15) B9(8-15) BC(8-15) BD(8-15)
+                    const __m512i rhs_mat_2367ABEF_1 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(rhs_raw_mat_2367ABEF_1, m4bexpanded)); //B2(8-15) B3(8-15) B6(8-15) B7(8-15) BA(8-15) BB(8-15) BE(8-15) BF(8-15)
+
+                    const __m512i rhs_mat_014589CD_2 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(_mm512_srli_epi16(rhs_raw_mat_014589CD_0, 4), m4bexpanded)); //B0(16-23) B1(16-23) B4(16-23) B5(16-23) B8(16-23) B9(16-23) BC(16-23) BD(16-23)
+                    const __m512i rhs_mat_2367ABEF_2 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(_mm512_srli_epi16(rhs_raw_mat_2367ABEF_0, 4), m4bexpanded)); //B2(16-23) B3(16-23) B6(16-23) B7(16-23) BA(16-23) BB(16-23) BE(16-23) BF(16-23)
+
+                    const __m512i rhs_mat_014589CD_3 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(_mm512_srli_epi16(rhs_raw_mat_014589CD_1, 4), m4bexpanded)); //B0(24-31) B1(24-31) B4(24-31) B5(24-31) B8(24-31) B9(24-31) BC(24-31) BD(24-31)
+                    const __m512i rhs_mat_2367ABEF_3 = _mm512_shuffle_epi8(signextendlutexpanded, _mm512_and_si512(_mm512_srli_epi16(rhs_raw_mat_2367ABEF_1, 4), m4bexpanded)); //B2(24-31) B3(24-31) B6(24-31) B7(24-31) BA(24-31) BB(24-31) BE(24-31) BF(24-31)
+
+                    // Shuffle pattern one - right side input
+                    const __m512i rhs_mat_014589CD_0_sp1 = _mm512_shuffle_epi32(rhs_mat_014589CD_0, 136); //B0(0-3) B1(0-3) B0(0-3) B1(0-3) B4(0-3) B5(0-3) B4(0-3) B5(0-3) B8(0-3) B9(0-3) B8(0-3) B9(0-3) BC(0-3) BD(0-3) BC(0-3) BD(0-3)
+                    const __m512i rhs_mat_2367ABEF_0_sp1 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_0, 136); //B2(0-3) B3(0-3) B2(0-3) B3(0-3) B6(0-3) B7(0-3) B6(0-3) B7(0-3) BA(0-3) BB(0-3) BA(0-3) BB(0-3) BE(0-3) BF(0-3) BE(0-3) BF(0-3)
+
+                    const __m512i rhs_mat_014589CD_1_sp1 = _mm512_shuffle_epi32(rhs_mat_014589CD_1, 136); //B0(8-11) B1(8-11) B0(8-11) B1(8-11) B4(8-11) B5(8-11) B4(8-11) B5(8-11) B8(8-11) B9(8-11) B8(8-11) B9(8-11) BC(8-11) BD(8-11) BC(8-11) BD(8-11)
+                    const __m512i rhs_mat_2367ABEF_1_sp1 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_1, 136); //B2(8-11) B3(8-11) B2(8-11) B3(8-11) B6(8-11) B7(8-11) B6(8-11) B7(8-11) BA(8-11) BB(8-11) BA(8-11) BB(8-11) BE(8-11) BF(8-11) BE(8-11) BF(8-11)
+
+                    const __m512i rhs_mat_014589CD_2_sp1 = _mm512_shuffle_epi32(rhs_mat_014589CD_2, 136); //B0(16-19) B1(16-19) B0(16-19) B1(16-19) B4(16-19) B5(16-19) B4(16-19) B5(16-19) B8(16-19) B9(16-19) B8(16-19) B9(16-19) BC(16-19) BD(16-19) BC(16-19) BD(16-19)
+                    const __m512i rhs_mat_2367ABEF_2_sp1 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_2, 136); //B2(16-19) B3(16-19) B2(16-19) B3(16-19) B6(16-19) B7(16-19) B6(16-19) B7(16-19) BA(16-19) BB(16-19) BA(16-19) BB(16-19) BE(16-19) BF(16-19) BE(16-19) BF(16-19)
+
+                    const __m512i rhs_mat_014589CD_3_sp1 = _mm512_shuffle_epi32(rhs_mat_014589CD_3, 136); //B0(24-27) B1(24-27) B0(24-27) B1(24-27) B4(24-27) B5(24-27) B4(24-27) B5(24-27) B8(24-27) B9(24-27) B8(24-27) B9(24-27) BC(24-27) BD(24-27) BC(24-27) BD(24-27)
+                    const __m512i rhs_mat_2367ABEF_3_sp1 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_3, 136); //B2(24-27) B3(24-27) B2(24-27) B3(24-27) B6(24-27) B7(24-27) B6(24-27) B7(24-27) BA(24-27) BB(24-27) BA(24-27) BB(24-27) BE(24-27) BF(24-27) BE(24-27) BF(24-27)
+
+                    // Shuffle pattern two - right side input
+
+                    const __m512i rhs_mat_014589CD_0_sp2 = _mm512_shuffle_epi32(rhs_mat_014589CD_0, 221); //B0(4-7) B1(4-7) B0(4-7) B1(4-7) B4(4-7) B5(4-7) B4(4-7) B5(4-7) B8(4-7) B9(4-7) B8(4-7) B9(4-7) BC(4-7) BD(4-7) BC(4-7) BD(4-7)
+                    const __m512i rhs_mat_2367ABEF_0_sp2 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_0, 221); //B2(4-7) B3(4-7) B2(4-7) B3(4-7) B6(4-7) B7(4-7) B6(4-7) B7(4-7) BA(4-7) BB(4-7) BA(4-7) BB(4-7) BE(4-7) BF(4-7) BE(4-7) BF(4-7)
+
+                    const __m512i rhs_mat_014589CD_1_sp2 = _mm512_shuffle_epi32(rhs_mat_014589CD_1, 221); //B0(12-15) B1(12-15) B0(12-15) B1(12-15) B4(12-15) B5(12-15) B4(12-15) B5(12-15) B8(12-15) B9(12-15) B8(12-15) B9(12-15) BC(12-15) BD(12-15) BC(12-15) BD(12-15)
+                    const __m512i rhs_mat_2367ABEF_1_sp2 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_1, 221); //B2(12-15) B3(12-15) B2(12-15) B3(12-15) B6(12-15) B7(12-15) B6(12-15) B7(12-15) BA(12-15) BB(12-15) BA(12-15) BB(12-15) BE(12-15) BF(12-15) BE(12-15) BF(12-15)
+
+                    const __m512i rhs_mat_014589CD_2_sp2 = _mm512_shuffle_epi32(rhs_mat_014589CD_2, 221); //B0(20-23) B1(20-23) B0(20-23) B1(20-23) B4(20-23) B5(20-23) B4(20-23) B5(20-23) B8(20-23) B9(20-23) B8(20-23) B9(20-23) BC(20-23) BD(20-23) BC(20-23) BD(20-23)
+                    const __m512i rhs_mat_2367ABEF_2_sp2 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_2, 221); //B2(20-23) B3(20-23) B2(20-23) B3(20-23) B6(20-23) B7(20-23) B6(20-23) B7(20-23) BA(20-23) BB(20-23) BA(20-23) BB(20-23) BE(20-23) BF(20-23) BE(20-23) BF(20-23)
+
+                    const __m512i rhs_mat_014589CD_3_sp2 = _mm512_shuffle_epi32(rhs_mat_014589CD_3, 221); //B0(28-31) B1(28-31) B0(28-31) B1(28-31) B4(28-31) B5(28-31) B4(28-31) B5(28-31) B8(28-31) B9(28-31) B8(28-31) B9(28-31) BC(28-31) BD(28-31) BC(28-31) BD(28-31)
+                    const __m512i rhs_mat_2367ABEF_3_sp2 = _mm512_shuffle_epi32(rhs_mat_2367ABEF_3, 221); //B2(28-31) B3(28-31) B2(28-31) B3(28-31) B6(28-31) B7(28-31) B6(28-31) B7(28-31) BA(28-31) BB(28-31) BA(28-31) BB(28-31) BE(28-31) BF(28-31) BE(28-31) BF(28-31)
+
+
+                    // Scale values - Load the weight scale values of two block_q4_0x8
+                    const __m512 col_scale_f32 = GGML_F32Cx8x2_LOAD(b_ptr_0[b].d, b_ptr_1[b].d);
+
+                    // Load the four block_q4_0 quantized values interleaved with each other in chunks of eight - A0,A1,A2,A3
+                    // Loaded as set of 128 bit vectors and repeated and stored into a 256 bit vector before again repeating into 512 bit vector
+                    __m256i lhs_mat_ymm_0123_0 = _mm256_loadu_si256((const __m256i *)((a_ptr[b].qs)));
+                    __m256i lhs_mat_ymm_01_0 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_0, lhs_mat_ymm_0123_0, 0);
+                    __m256i lhs_mat_ymm_23_0 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_0, lhs_mat_ymm_0123_0, 17);
+                    __m256i lhs_mat_ymm_0123_1 = _mm256_loadu_si256((const __m256i *)((a_ptr[b].qs + 32)));
+                    __m256i lhs_mat_ymm_01_1 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_1, lhs_mat_ymm_0123_1, 0);
+                    __m256i lhs_mat_ymm_23_1 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_1, lhs_mat_ymm_0123_1, 17);
+                    __m256i lhs_mat_ymm_0123_2 = _mm256_loadu_si256((const __m256i *)((a_ptr[b].qs + 64)));
+                    __m256i lhs_mat_ymm_01_2 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_2, lhs_mat_ymm_0123_2, 0);
+                    __m256i lhs_mat_ymm_23_2 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_2, lhs_mat_ymm_0123_2, 17);
+                    __m256i lhs_mat_ymm_0123_3 = _mm256_loadu_si256((const __m256i *)((a_ptr[b].qs + 96)));
+                    __m256i lhs_mat_ymm_01_3 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_3, lhs_mat_ymm_0123_3, 0);
+                    __m256i lhs_mat_ymm_23_3 = _mm256_permute2f128_si256(lhs_mat_ymm_0123_3, lhs_mat_ymm_0123_3, 17);
+
+                    __m512i lhs_mat_01_0 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_01_0), lhs_mat_ymm_01_0, 1);
+                    __m512i lhs_mat_23_0 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_23_0), lhs_mat_ymm_23_0, 1);
+                    __m512i lhs_mat_01_1 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_01_1), lhs_mat_ymm_01_1, 1);
+                    __m512i lhs_mat_23_1 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_23_1), lhs_mat_ymm_23_1, 1);
+                    __m512i lhs_mat_01_2 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_01_2), lhs_mat_ymm_01_2, 1);
+                    __m512i lhs_mat_23_2 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_23_2), lhs_mat_ymm_23_2, 1);
+                    __m512i lhs_mat_01_3 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_01_3), lhs_mat_ymm_01_3, 1);
+                    __m512i lhs_mat_23_3 = _mm512_inserti32x8(_mm512_castsi256_si512(lhs_mat_ymm_23_3), lhs_mat_ymm_23_3, 1);
+
+                    // Shuffle pattern one - left side input
+
+                    const __m512i lhs_mat_01_0_sp1 = _mm512_shuffle_epi32(lhs_mat_01_0, 160);  //A0(0-3) A0(0-3) A1(0-3) A1(0-3) A0(0-3) A0(0-3) A1(0-3) A1(0-3) A0(0-3) A0(0-3) A1(0-3) A1(0-3) A0(0-3) A0(0-3) A1(0-3) A1(0-3)
+                    const __m512i lhs_mat_23_0_sp1 = _mm512_shuffle_epi32(lhs_mat_23_0, 160);  //A2(0-3) A2(0-3) A3(0-3) A3(0-3) A2(0-3) A2(0-3) A3(0-3) A3(0-3) A2(0-3) A2(0-3) A3(0-3) A3(0-3) A2(0-3) A2(0-3) A3(0-3) A3(0-3)
+
+                    const __m512i lhs_mat_01_1_sp1 = _mm512_shuffle_epi32(lhs_mat_01_1, 160);  //A0(8-11) A0(8-11) A1(8-11) A1(8-11) A0(8-11) A0(8-11) A1(8-11) A1(8-11) A0(8-11) A0(8-11) A1(8-11) A1(8-11) A0(8-11) A0(8-11) A1(8-11) A1(8-11)
+                    const __m512i lhs_mat_23_1_sp1 = _mm512_shuffle_epi32(lhs_mat_23_1, 160);  //A2(8-11) A2(8-11) A3(8-11) A3(8-11) A2(8-11) A2(8-11) A3(8-11) A3(8-11) A2(8-11) A2(8-11) A3(8-11) A3(8-11) A2(8-11) A2(8-11) A3(8-11) A3(8-11)
+
+                    const __m512i lhs_mat_01_2_sp1 = _mm512_shuffle_epi32(lhs_mat_01_2, 160);  //A0(16-19) A0(16-19) A1(16-19) A1(16-19) A0(16-19) A0(16-19) A1(16-19) A1(16-19) A0(16-19) A0(16-19) A1(16-19) A1(16-19) A0(16-19) A0(16-19) A1(16-19) A1(16-19)
+                    const __m512i lhs_mat_23_2_sp1 = _mm512_shuffle_epi32(lhs_mat_23_2, 160);  //A2(16-19) A2(16-19) A3(16-19) A3(16-19) A2(16-19) A2(16-19) A3(16-19) A3(16-19) A2(16-19) A2(16-19) A3(16-19) A3(16-19) A2(16-19) A2(16-19) A3(16-19) A3(16-19)
+
+                    const __m512i lhs_mat_01_3_sp1 = _mm512_shuffle_epi32(lhs_mat_01_3, 160);  //A0(24-27) A0(24-27) A1(24-27) A1(24-27) A0(24-27) A0(24-27) A1(24-27) A1(24-27) A0(24-27) A0(24-27) A1(24-27) A1(24-27) A0(24-27) A0(24-27) A1(24-27) A1(24-27)
+                    const __m512i lhs_mat_23_3_sp1 = _mm512_shuffle_epi32(lhs_mat_23_3, 160);  //A2(24-27) A2(24-27) A3(24-27) A3(24-27) A2(24-27) A2(24-27) A3(24-27) A3(24-27) A2(24-27) A2(24-27) A3(24-27) A3(24-27) A2(24-27) A2(24-27) A3(24-27) A3(24-27)
+
+                    // Shuffle pattern two - left side input
+
+                    const __m512i lhs_mat_01_0_sp2 = _mm512_shuffle_epi32(lhs_mat_01_0, 245);  //A0(4-7) A0(4-7) A1(4-7) A1(4-7) A0(4-7) A0(4-7) A1(4-7) A1(4-7) A0(4-7) A0(4-7) A1(4-7) A1(4-7) A0(4-7) A0(4-7) A1(4-7) A1(4-7)
+                    const __m512i lhs_mat_23_0_sp2 = _mm512_shuffle_epi32(lhs_mat_23_0, 245);  //A2(4-7) A2(4-7) A3(4-7) A3(4-7) A2(4-7) A2(4-7) A3(4-7) A3(4-7) A2(4-7) A2(4-7) A3(4-7) A3(4-7) A2(4-7) A2(4-7) A3(4-7) A3(4-7)
+
+                    const __m512i lhs_mat_01_1_sp2 = _mm512_shuffle_epi32(lhs_mat_01_1, 245);  //A0(12-15) A0(12-15) A1(12-15) A1(12-15) A0(12-15) A0(12-15) A1(12-15) A1(12-15) A0(12-15) A0(12-15) A1(12-15) A1(12-15) A0(12-15) A0(12-15) A1(12-15) A1(12-15)
+                    const __m512i lhs_mat_23_1_sp2 = _mm512_shuffle_epi32(lhs_mat_23_1, 245);  //A2(12-15) A2(12-15) A3(12-15) A3(12-15) A2(12-15) A2(12-15) A3(12-15) A3(12-15) A2(12-15) A2(12-15) A3(12-15) A3(12-15) A2(12-15) A2(12-15) A3(12-15) A3(12-15)
+
+                    const __m512i lhs_mat_01_2_sp2 = _mm512_shuffle_epi32(lhs_mat_01_2, 245);  //A0(20-23) A0(20-23) A1(20-23) A1(20-23) A0(20-23) A0(20-23) A1(20-23) A1(20-23) A0(20-23) A0(20-23) A1(20-23) A1(20-23) A0(20-23) A0(20-23) A1(20-23) A1(20-23)
+                    const __m512i lhs_mat_23_2_sp2 = _mm512_shuffle_epi32(lhs_mat_23_2, 245);  //A2(20-23) A2(20-23) A3(20-23) A3(20-23) A2(20-23) A2(20-23) A3(20-23) A3(20-23) A2(20-23) A2(20-23) A3(20-23) A3(20-23) A2(20-23) A2(20-23) A3(20-23) A3(20-23)
+
+                    const __m512i lhs_mat_01_3_sp2 = _mm512_shuffle_epi32(lhs_mat_01_3, 245);  //A0(28-31) A0(28-31) A1(28-31) A1(28-31) A0(28-31) A0(28-31) A1(28-31) A1(28-31) A0(28-31) A0(28-31) A1(28-31) A1(28-31) A0(28-31) A0(28-31) A1(28-31) A1(28-31)
+                    const __m512i lhs_mat_23_3_sp2 = _mm512_shuffle_epi32(lhs_mat_23_3, 245);  //A2(28-31) A2(28-31) A3(28-31) A3(28-31) A2(28-31) A2(28-31) A3(28-31) A3(28-31) A2(28-31) A2(28-31) A3(28-31) A3(28-31) A2(28-31) A2(28-31) A3(28-31) A3(28-31)
+
+                    // The values arranged in shuffle patterns are operated with dot product operation within 32 bit lane i.e corresponding bytes and multiplied and added into 32 bit integers within 32 bit lane
+                    // Resembles MMLAs into 2x2 matrices in ARM Version
+                    __m512i iacc_mat_00_sp1 =
+                        _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_01_3_sp1, rhs_mat_014589CD_3_sp1), mul_sum_i8_pairs_int32x16(lhs_mat_01_2_sp1, rhs_mat_014589CD_2_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_01_1_sp1, rhs_mat_014589CD_1_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_01_0_sp1, rhs_mat_014589CD_0_sp1));
+                    __m512i iacc_mat_01_sp1 =
+                        _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_01_3_sp1, rhs_mat_2367ABEF_3_sp1), mul_sum_i8_pairs_int32x16(lhs_mat_01_2_sp1, rhs_mat_2367ABEF_2_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_01_1_sp1, rhs_mat_2367ABEF_1_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_01_0_sp1, rhs_mat_2367ABEF_0_sp1));
+                    __m512i iacc_mat_10_sp1 =
+                        _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_23_3_sp1, rhs_mat_014589CD_3_sp1), mul_sum_i8_pairs_int32x16(lhs_mat_23_2_sp1, rhs_mat_014589CD_2_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_23_1_sp1, rhs_mat_014589CD_1_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_23_0_sp1, rhs_mat_014589CD_0_sp1));
+                    __m512i iacc_mat_11_sp1 =
+                        _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_23_3_sp1, rhs_mat_2367ABEF_3_sp1), mul_sum_i8_pairs_int32x16(lhs_mat_23_2_sp1, rhs_mat_2367ABEF_2_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_23_1_sp1, rhs_mat_2367ABEF_1_sp1)), mul_sum_i8_pairs_int32x16(lhs_mat_23_0_sp1, rhs_mat_2367ABEF_0_sp1));
+                    __m512i iacc_mat_00_sp2 =
+                        _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_01_3_sp2, rhs_mat_014589CD_3_sp2), mul_sum_i8_pairs_int32x16(lhs_mat_01_2_sp2, rhs_mat_014589CD_2_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_01_1_sp2, rhs_mat_014589CD_1_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_01_0_sp2, rhs_mat_014589CD_0_sp2));
+                    __m512i iacc_mat_01_sp2 =
+                        _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_01_3_sp2, rhs_mat_2367ABEF_3_sp2), mul_sum_i8_pairs_int32x16(lhs_mat_01_2_sp2, rhs_mat_2367ABEF_2_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_01_1_sp2, rhs_mat_2367ABEF_1_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_01_0_sp2, rhs_mat_2367ABEF_0_sp2));
+                    __m512i iacc_mat_10_sp2 =
+                        _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_23_3_sp2, rhs_mat_014589CD_3_sp2), mul_sum_i8_pairs_int32x16(lhs_mat_23_2_sp2, rhs_mat_014589CD_2_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_23_1_sp2, rhs_mat_014589CD_1_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_23_0_sp2, rhs_mat_014589CD_0_sp2));
+                    __m512i iacc_mat_11_sp2 =
+                        _mm512_add_epi32(_mm512_add_epi32(_mm512_add_epi32(mul_sum_i8_pairs_int32x16(lhs_mat_23_3_sp2, rhs_mat_2367ABEF_3_sp2), mul_sum_i8_pairs_int32x16(lhs_mat_23_2_sp2, rhs_mat_2367ABEF_2_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_23_1_sp2, rhs_mat_2367ABEF_1_sp2)), mul_sum_i8_pairs_int32x16(lhs_mat_23_0_sp2, rhs_mat_2367ABEF_0_sp2));
+
+                    // Output of both shuffle patterns are added in order to sum dot product outputs of all 32 values in block
+                    __m512i iacc_mat_00 = _mm512_add_epi32(iacc_mat_00_sp1, iacc_mat_00_sp2);
+                    __m512i iacc_mat_01 = _mm512_add_epi32(iacc_mat_01_sp1, iacc_mat_01_sp2);
+                    __m512i iacc_mat_10 = _mm512_add_epi32(iacc_mat_10_sp1, iacc_mat_10_sp2);
+                    __m512i iacc_mat_11 = _mm512_add_epi32(iacc_mat_11_sp1, iacc_mat_11_sp2);
+
+
+                    // Straighten out to make 4 row vectors
+                    __m512i iacc_row_0 = _mm512_mask_blend_epi32(0xCCCC, iacc_mat_00, _mm512_shuffle_epi32(iacc_mat_01, 78));
+                    __m512i iacc_row_1 = _mm512_mask_blend_epi32(0xCCCC, _mm512_shuffle_epi32(iacc_mat_00, 78), iacc_mat_01);
+                    __m512i iacc_row_2 = _mm512_mask_blend_epi32(0xCCCC, iacc_mat_10, _mm512_shuffle_epi32(iacc_mat_11, 78));
+                    __m512i iacc_row_3 = _mm512_mask_blend_epi32(0xCCCC, _mm512_shuffle_epi32(iacc_mat_10, 78), iacc_mat_11);
+
+                    // Load the scale(d) values for all the 4 Q8_0 blocks and repeat it across lanes
+                    const __m128i row_scale_f16 = _mm_shuffle_epi32(_mm_maskload_epi32((int const*)(a_ptr[b].d), loadMask), 68);
+                    const __m512 row_scale_f32 = GGML_F32Cx16_REPEAT_LOAD(row_scale_f16);
+
+                    // Multiply with appropiate scales and accumulate
+                    acc_rows[0] = _mm512_fmadd_ps(_mm512_cvtepi32_ps(iacc_row_0), _mm512_mul_ps(col_scale_f32, _mm512_shuffle_ps(row_scale_f32, row_scale_f32, 0)),   acc_rows[0]);
+                    acc_rows[1] = _mm512_fmadd_ps(_mm512_cvtepi32_ps(iacc_row_1), _mm512_mul_ps(col_scale_f32, _mm512_shuffle_ps(row_scale_f32, row_scale_f32, 85)),  acc_rows[1]);
+                    acc_rows[2] = _mm512_fmadd_ps(_mm512_cvtepi32_ps(iacc_row_2), _mm512_mul_ps(col_scale_f32, _mm512_shuffle_ps(row_scale_f32, row_scale_f32, 170)), acc_rows[2]);
+                    acc_rows[3] = _mm512_fmadd_ps(_mm512_cvtepi32_ps(iacc_row_3), _mm512_mul_ps(col_scale_f32, _mm512_shuffle_ps(row_scale_f32, row_scale_f32, 255)), acc_rows[3]);
+                }
+
+                // Store the accumulated values
+                for (int i = 0; i < 4; i++) {
+                    _mm512_storeu_ps((float *)(s + ((y * 4 + i) * bs + x * 8)), acc_rows[i]);
+                }
+            }
+        }
+        if (anc != nc) {
+            xstart = anc/8;
+            y = 0;
+        }
+    #endif // __AVX512F__
 
         // Take group of four block_q8_0x4 structures at each pass of the loop and perform dot product operation
-    int anr = nr - nr %16; // Used to align nr with boundary of 16
 
         for (; y < anr / 4; y += 4) {
             const block_q8_0x4 * a_ptrs[4];
@@ -2435,7 +2850,7 @@ void ggml_gemm_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
             }
 
             // Take group of eight block_q4_0x8 structures at each pass of the loop and perform dot product operation
-        for (int64_t x = 0; x < nc / 8; x++) {
+            for (int64_t x = xstart; x < nc / 8; x++) {
 
                 const block_q4_0x8 * b_ptr = b_ptr_start + (x * b_nb);
 
@@ -2547,21 +2962,21 @@ void ggml_gemm_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
                         // The values arranged in shuffle patterns are operated with dot product operation within 32 bit lane i.e corresponding bytes and multiplied and added into 32 bit integers within 32 bit lane
                         // Resembles MMLAs into 2x2 matrices in ARM Version
                         __m256i iacc_mat_00_sp1 =
-                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_01_3_sp1, rhs_mat_0145_3_sp1), mul_sum_i8_pairs_int(lhs_mat_01_2_sp1, rhs_mat_0145_2_sp1)), mul_sum_i8_pairs_int(lhs_mat_01_1_sp1, rhs_mat_0145_1_sp1)), mul_sum_i8_pairs_int(lhs_mat_01_0_sp1, rhs_mat_0145_0_sp1));
+                            _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_01_3_sp1, rhs_mat_0145_3_sp1), mul_sum_i8_pairs_int32x8(lhs_mat_01_2_sp1, rhs_mat_0145_2_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_01_1_sp1, rhs_mat_0145_1_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_01_0_sp1, rhs_mat_0145_0_sp1));
                         __m256i iacc_mat_01_sp1 =
-                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_01_3_sp1, rhs_mat_2367_3_sp1), mul_sum_i8_pairs_int(lhs_mat_01_2_sp1, rhs_mat_2367_2_sp1)), mul_sum_i8_pairs_int(lhs_mat_01_1_sp1, rhs_mat_2367_1_sp1)), mul_sum_i8_pairs_int(lhs_mat_01_0_sp1, rhs_mat_2367_0_sp1));
+                            _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_01_3_sp1, rhs_mat_2367_3_sp1), mul_sum_i8_pairs_int32x8(lhs_mat_01_2_sp1, rhs_mat_2367_2_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_01_1_sp1, rhs_mat_2367_1_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_01_0_sp1, rhs_mat_2367_0_sp1));
                         __m256i iacc_mat_10_sp1 =
-                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_23_3_sp1, rhs_mat_0145_3_sp1), mul_sum_i8_pairs_int(lhs_mat_23_2_sp1, rhs_mat_0145_2_sp1)), mul_sum_i8_pairs_int(lhs_mat_23_1_sp1, rhs_mat_0145_1_sp1)), mul_sum_i8_pairs_int(lhs_mat_23_0_sp1, rhs_mat_0145_0_sp1));
+                            _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_23_3_sp1, rhs_mat_0145_3_sp1), mul_sum_i8_pairs_int32x8(lhs_mat_23_2_sp1, rhs_mat_0145_2_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_23_1_sp1, rhs_mat_0145_1_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_23_0_sp1, rhs_mat_0145_0_sp1));
                         __m256i iacc_mat_11_sp1 =
-                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_23_3_sp1, rhs_mat_2367_3_sp1), mul_sum_i8_pairs_int(lhs_mat_23_2_sp1, rhs_mat_2367_2_sp1)), mul_sum_i8_pairs_int(lhs_mat_23_1_sp1, rhs_mat_2367_1_sp1)), mul_sum_i8_pairs_int(lhs_mat_23_0_sp1, rhs_mat_2367_0_sp1));
+                            _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_23_3_sp1, rhs_mat_2367_3_sp1), mul_sum_i8_pairs_int32x8(lhs_mat_23_2_sp1, rhs_mat_2367_2_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_23_1_sp1, rhs_mat_2367_1_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_23_0_sp1, rhs_mat_2367_0_sp1));
                         __m256i iacc_mat_00_sp2 =
-                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_01_3_sp2, rhs_mat_0145_3_sp2), mul_sum_i8_pairs_int(lhs_mat_01_2_sp2, rhs_mat_0145_2_sp2)), mul_sum_i8_pairs_int(lhs_mat_01_1_sp2, rhs_mat_0145_1_sp2)), mul_sum_i8_pairs_int(lhs_mat_01_0_sp2, rhs_mat_0145_0_sp2));
+                            _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_01_3_sp2, rhs_mat_0145_3_sp2), mul_sum_i8_pairs_int32x8(lhs_mat_01_2_sp2, rhs_mat_0145_2_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_01_1_sp2, rhs_mat_0145_1_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_01_0_sp2, rhs_mat_0145_0_sp2));
                         __m256i iacc_mat_01_sp2 =
-                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_01_3_sp2, rhs_mat_2367_3_sp2), mul_sum_i8_pairs_int(lhs_mat_01_2_sp2, rhs_mat_2367_2_sp2)), mul_sum_i8_pairs_int(lhs_mat_01_1_sp2, rhs_mat_2367_1_sp2)), mul_sum_i8_pairs_int(lhs_mat_01_0_sp2, rhs_mat_2367_0_sp2));
+                            _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_01_3_sp2, rhs_mat_2367_3_sp2), mul_sum_i8_pairs_int32x8(lhs_mat_01_2_sp2, rhs_mat_2367_2_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_01_1_sp2, rhs_mat_2367_1_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_01_0_sp2, rhs_mat_2367_0_sp2));
                         __m256i iacc_mat_10_sp2 =
-                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_23_3_sp2, rhs_mat_0145_3_sp2), mul_sum_i8_pairs_int(lhs_mat_23_2_sp2, rhs_mat_0145_2_sp2)), mul_sum_i8_pairs_int(lhs_mat_23_1_sp2, rhs_mat_0145_1_sp2)), mul_sum_i8_pairs_int(lhs_mat_23_0_sp2, rhs_mat_0145_0_sp2));
+                            _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_23_3_sp2, rhs_mat_0145_3_sp2), mul_sum_i8_pairs_int32x8(lhs_mat_23_2_sp2, rhs_mat_0145_2_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_23_1_sp2, rhs_mat_0145_1_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_23_0_sp2, rhs_mat_0145_0_sp2));
                         __m256i iacc_mat_11_sp2 =
-                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_23_3_sp2, rhs_mat_2367_3_sp2), mul_sum_i8_pairs_int(lhs_mat_23_2_sp2, rhs_mat_2367_2_sp2)), mul_sum_i8_pairs_int(lhs_mat_23_1_sp2, rhs_mat_2367_1_sp2)), mul_sum_i8_pairs_int(lhs_mat_23_0_sp2, rhs_mat_2367_0_sp2));
+                            _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_23_3_sp2, rhs_mat_2367_3_sp2), mul_sum_i8_pairs_int32x8(lhs_mat_23_2_sp2, rhs_mat_2367_2_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_23_1_sp2, rhs_mat_2367_1_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_23_0_sp2, rhs_mat_2367_0_sp2));
 
                         // Output of both shuffle patterns are added in order to sum dot product outputs of all 32 values in block
                         __m256i iacc_mat_00 = _mm256_add_epi32(iacc_mat_00_sp1, iacc_mat_00_sp2);
@@ -2599,7 +3014,7 @@ void ggml_gemm_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
             const block_q8_0x4 * a_ptr = a_ptr_start + (y * nb);
 
             // Load the eight block_q4_0 quantized values interleaved with each other in chunks of eight - B0,B1 ....B6,B7
-        for (int64_t x = 0; x < nc / 8; x++) {
+            for (int64_t x = xstart; x < nc / 8; x++) {
 
                 const block_q4_0x8 * b_ptr = b_ptr_start + (x * b_nb);
 
@@ -2711,21 +3126,21 @@ void ggml_gemm_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
                     // The values arranged in shuffle patterns are operated with dot product operation within 32 bit lane i.e corresponding bytes and multiplied and added into 32 bit integers within 32 bit lane
                     // Resembles MMLAs into 2x2 matrices in ARM Version
                     __m256i iacc_mat_00_sp1 =
-                    _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_01_3_sp1, rhs_mat_0145_3_sp1), mul_sum_i8_pairs_int(lhs_mat_01_2_sp1, rhs_mat_0145_2_sp1)), mul_sum_i8_pairs_int(lhs_mat_01_1_sp1, rhs_mat_0145_1_sp1)), mul_sum_i8_pairs_int(lhs_mat_01_0_sp1, rhs_mat_0145_0_sp1));
+                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_01_3_sp1, rhs_mat_0145_3_sp1), mul_sum_i8_pairs_int32x8(lhs_mat_01_2_sp1, rhs_mat_0145_2_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_01_1_sp1, rhs_mat_0145_1_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_01_0_sp1, rhs_mat_0145_0_sp1));
                     __m256i iacc_mat_01_sp1 =
-                    _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_01_3_sp1, rhs_mat_2367_3_sp1), mul_sum_i8_pairs_int(lhs_mat_01_2_sp1, rhs_mat_2367_2_sp1)), mul_sum_i8_pairs_int(lhs_mat_01_1_sp1, rhs_mat_2367_1_sp1)), mul_sum_i8_pairs_int(lhs_mat_01_0_sp1, rhs_mat_2367_0_sp1));
+                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_01_3_sp1, rhs_mat_2367_3_sp1), mul_sum_i8_pairs_int32x8(lhs_mat_01_2_sp1, rhs_mat_2367_2_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_01_1_sp1, rhs_mat_2367_1_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_01_0_sp1, rhs_mat_2367_0_sp1));
                     __m256i iacc_mat_10_sp1 =
-                    _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_23_3_sp1, rhs_mat_0145_3_sp1), mul_sum_i8_pairs_int(lhs_mat_23_2_sp1, rhs_mat_0145_2_sp1)), mul_sum_i8_pairs_int(lhs_mat_23_1_sp1, rhs_mat_0145_1_sp1)), mul_sum_i8_pairs_int(lhs_mat_23_0_sp1, rhs_mat_0145_0_sp1));
+                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_23_3_sp1, rhs_mat_0145_3_sp1), mul_sum_i8_pairs_int32x8(lhs_mat_23_2_sp1, rhs_mat_0145_2_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_23_1_sp1, rhs_mat_0145_1_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_23_0_sp1, rhs_mat_0145_0_sp1));
                     __m256i iacc_mat_11_sp1 =
-                    _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_23_3_sp1, rhs_mat_2367_3_sp1), mul_sum_i8_pairs_int(lhs_mat_23_2_sp1, rhs_mat_2367_2_sp1)), mul_sum_i8_pairs_int(lhs_mat_23_1_sp1, rhs_mat_2367_1_sp1)), mul_sum_i8_pairs_int(lhs_mat_23_0_sp1, rhs_mat_2367_0_sp1));
+                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_23_3_sp1, rhs_mat_2367_3_sp1), mul_sum_i8_pairs_int32x8(lhs_mat_23_2_sp1, rhs_mat_2367_2_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_23_1_sp1, rhs_mat_2367_1_sp1)), mul_sum_i8_pairs_int32x8(lhs_mat_23_0_sp1, rhs_mat_2367_0_sp1));
                     __m256i iacc_mat_00_sp2 =
-                    _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_01_3_sp2, rhs_mat_0145_3_sp2), mul_sum_i8_pairs_int(lhs_mat_01_2_sp2, rhs_mat_0145_2_sp2)), mul_sum_i8_pairs_int(lhs_mat_01_1_sp2, rhs_mat_0145_1_sp2)), mul_sum_i8_pairs_int(lhs_mat_01_0_sp2, rhs_mat_0145_0_sp2));
+                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_01_3_sp2, rhs_mat_0145_3_sp2), mul_sum_i8_pairs_int32x8(lhs_mat_01_2_sp2, rhs_mat_0145_2_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_01_1_sp2, rhs_mat_0145_1_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_01_0_sp2, rhs_mat_0145_0_sp2));
                     __m256i iacc_mat_01_sp2 =
-                    _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_01_3_sp2, rhs_mat_2367_3_sp2), mul_sum_i8_pairs_int(lhs_mat_01_2_sp2, rhs_mat_2367_2_sp2)), mul_sum_i8_pairs_int(lhs_mat_01_1_sp2, rhs_mat_2367_1_sp2)), mul_sum_i8_pairs_int(lhs_mat_01_0_sp2, rhs_mat_2367_0_sp2));
+                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_01_3_sp2, rhs_mat_2367_3_sp2), mul_sum_i8_pairs_int32x8(lhs_mat_01_2_sp2, rhs_mat_2367_2_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_01_1_sp2, rhs_mat_2367_1_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_01_0_sp2, rhs_mat_2367_0_sp2));
                     __m256i iacc_mat_10_sp2 =
-                    _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_23_3_sp2, rhs_mat_0145_3_sp2), mul_sum_i8_pairs_int(lhs_mat_23_2_sp2, rhs_mat_0145_2_sp2)), mul_sum_i8_pairs_int(lhs_mat_23_1_sp2, rhs_mat_0145_1_sp2)), mul_sum_i8_pairs_int(lhs_mat_23_0_sp2, rhs_mat_0145_0_sp2));
+                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_23_3_sp2, rhs_mat_0145_3_sp2), mul_sum_i8_pairs_int32x8(lhs_mat_23_2_sp2, rhs_mat_0145_2_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_23_1_sp2, rhs_mat_0145_1_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_23_0_sp2, rhs_mat_0145_0_sp2));
                     __m256i iacc_mat_11_sp2 =
-                    _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int(lhs_mat_23_3_sp2, rhs_mat_2367_3_sp2), mul_sum_i8_pairs_int(lhs_mat_23_2_sp2, rhs_mat_2367_2_sp2)), mul_sum_i8_pairs_int(lhs_mat_23_1_sp2, rhs_mat_2367_1_sp2)), mul_sum_i8_pairs_int(lhs_mat_23_0_sp2, rhs_mat_2367_0_sp2));
+                        _mm256_add_epi32(_mm256_add_epi32(_mm256_add_epi32(mul_sum_i8_pairs_int32x8(lhs_mat_23_3_sp2, rhs_mat_2367_3_sp2), mul_sum_i8_pairs_int32x8(lhs_mat_23_2_sp2, rhs_mat_2367_2_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_23_1_sp2, rhs_mat_2367_1_sp2)), mul_sum_i8_pairs_int32x8(lhs_mat_23_0_sp2, rhs_mat_2367_0_sp2));
 
                     // Output of both shuffle patterns are added in order to sum dot product outputs of all 32 values in block
                     __m256i iacc_mat_00 = _mm256_add_epi32(iacc_mat_00_sp1, iacc_mat_00_sp2);
@@ -2756,7 +3171,9 @@ void ggml_gemm_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
                 }
             }
         }
-#else
+        return;
+    }
+#endif // #if ! ((defined(_MSC_VER)) && ! defined(__clang__)) && defined(__aarch64__)
     float sumf[4][8];
     int sumi;
 
@@ -2789,5 +3206,4 @@ void ggml_gemm_q4_0_8x8_q8_0(int n, float * restrict s, size_t bs, const void *
             }
         }
     }
-#endif
 }

ggml/src/ggml-alloc.c

@@ -294,6 +294,12 @@ static void ggml_dyn_tallocr_reset(struct ggml_dyn_tallocr * alloc) {
     alloc->free_blocks[0].offset = 0;
     alloc->free_blocks[0].size = SIZE_MAX/2; // restrict maximum size of a measure allocator to half size_t max to avoid overflows
     alloc->max_size = 0;
+
+#ifdef GGML_ALLOCATOR_DEBUG
+    for (int i = 0; i < 1024; i++) {
+        alloc->allocated_tensors[i].tensor = NULL;
+    }
+#endif
 }
 
 static struct ggml_dyn_tallocr * ggml_dyn_tallocr_new(size_t alignment) {

ggml/src/ggml-backend-impl.h

@@ -39,9 +39,10 @@ extern "C" {
 
     struct ggml_backend_buffer_i {
         const char * (*GGML_CALL get_name)      (ggml_backend_buffer_t buffer);
-        void         (*GGML_CALL free_buffer)(ggml_backend_buffer_t buffer);
+        void         (*GGML_CALL free_buffer)   (ggml_backend_buffer_t buffer);
         void *       (*GGML_CALL get_base)      (ggml_backend_buffer_t buffer);
-        void         (*GGML_CALL init_tensor)(ggml_backend_buffer_t buffer, struct ggml_tensor * tensor);
+        void         (*GGML_CALL init_tensor)   (ggml_backend_buffer_t buffer, struct ggml_tensor * tensor);
+        void         (*GGML_CALL memset_tensor) (ggml_backend_buffer_t buffer,       struct ggml_tensor * tensor,     uint8_t value, size_t offset, size_t size);
         void         (*GGML_CALL set_tensor)    (ggml_backend_buffer_t buffer,       struct ggml_tensor * tensor, const void * data, size_t offset, size_t size);
         void         (*GGML_CALL get_tensor)    (ggml_backend_buffer_t buffer, const struct ggml_tensor * tensor,       void * data, size_t offset, size_t size);
         bool         (*GGML_CALL cpy_tensor)    (ggml_backend_buffer_t buffer, const struct ggml_tensor * src, struct ggml_tensor * dst); // dst is in the buffer, src may be in any buffer

ggml/src/ggml-backend.c

@@ -246,6 +246,22 @@ GGML_CALL void ggml_backend_tensor_get(const struct ggml_tensor * tensor, void *
     buf->iface.get_tensor(buf, tensor, data, offset, size);
 }
 
+GGML_API GGML_CALL void ggml_backend_tensor_memset(struct ggml_tensor * tensor, uint8_t value, size_t offset, size_t size) {
+    ggml_backend_buffer_t buf = tensor->view_src ? tensor->view_src->buffer : tensor->buffer;
+
+    GGML_ASSERT(buf != NULL && "tensor buffer not set");
+    GGML_ASSERT(tensor->data != NULL && "tensor not allocated");
+    GGML_ASSERT(offset + size <= ggml_nbytes(tensor) && "tensor write out of bounds");
+
+    if (!size) {
+        return;
+    }
+
+    GGML_ASSERT(buf->iface.memset_tensor != NULL && "memset not supported by backend buffer");
+
+    buf->iface.memset_tensor(buf, tensor, value, offset, size);
+}
+
 void ggml_backend_synchronize(ggml_backend_t backend) {
     if (backend->iface.synchronize == NULL) {
         return;
@@ -569,6 +585,12 @@ GGML_CALL static void ggml_backend_cpu_buffer_free_buffer(ggml_backend_buffer_t
     free(buffer->context);
 }
 
+GGML_CALL static void ggml_backend_cpu_buffer_memset_tensor(ggml_backend_buffer_t buffer, struct ggml_tensor * tensor, uint8_t value, size_t offset, size_t size) {
+    memset((char *)tensor->data + offset, value, size);
+
+    GGML_UNUSED(buffer);
+}
+
 GGML_CALL static void ggml_backend_cpu_buffer_set_tensor(ggml_backend_buffer_t buffer, struct ggml_tensor * tensor, const void * data, size_t offset, size_t size) {
     memcpy((char *)tensor->data + offset, data, size);
 
@@ -600,6 +622,7 @@ static struct ggml_backend_buffer_i cpu_backend_buffer_i = {
     /* .free_buffer     = */ ggml_backend_cpu_buffer_free_buffer,
     /* .get_base        = */ ggml_backend_cpu_buffer_get_base,
     /* .init_tensor     = */ NULL, // no initialization required
+    /* .memset_tensor   = */ ggml_backend_cpu_buffer_memset_tensor,
     /* .set_tensor      = */ ggml_backend_cpu_buffer_set_tensor,
     /* .get_tensor      = */ ggml_backend_cpu_buffer_get_tensor,
     /* .cpy_tensor      = */ ggml_backend_cpu_buffer_cpy_tensor,
@@ -613,6 +636,7 @@ static struct ggml_backend_buffer_i cpu_backend_buffer_i_from_ptr = {
     /* .free_buffer     = */ NULL, // ptr is not owned by the buffer, so it does not need to be freed
     /* .get_base        = */ ggml_backend_cpu_buffer_get_base,
     /* .init_tensor     = */ NULL, // no initialization required
+    /* .memset_tensor   = */ ggml_backend_cpu_buffer_memset_tensor,
     /* .set_tensor      = */ ggml_backend_cpu_buffer_set_tensor,
     /* .get_tensor      = */ ggml_backend_cpu_buffer_get_tensor,
     /* .cpy_tensor      = */ ggml_backend_cpu_buffer_cpy_tensor,
@@ -980,6 +1004,7 @@ static struct ggml_backend_buffer_i ggml_backend_multi_buffer_context_interface(
         /* .free_buffer     = */ ggml_backend_multi_buffer_free_buffer,
         /* .get_base        = */ NULL,
         /* .init_tensor     = */ NULL,
+        /* .memset_tensor   = */ NULL,
         /* .set_tensor      = */ NULL,
         /* .get_tensor      = */ NULL,
         /* .cpy_tensor      = */ NULL,

ggml/src/ggml-cann.cpp

@@ -1037,6 +1037,7 @@ static ggml_backend_buffer_i ggml_backend_cann_buffer_interface = {
     /* .free_buffer     = */ ggml_backend_cann_buffer_free_buffer,
     /* .get_base        = */ ggml_backend_cann_buffer_get_base,
     /* .init_tensor     = */ ggml_backend_cann_buffer_init_tensor,
+    /* .memset_tensor   = */ NULL,
     /* .set_tensor      = */ ggml_backend_cann_buffer_set_tensor,
     /* .get_tensor      = */ ggml_backend_cann_buffer_get_tensor,
     /* .cpy_tensor      = */ ggml_backend_cann_buffer_cpy_tensor,

ggml/src/ggml-cann/common.h

@@ -227,6 +227,7 @@ struct ggml_backend_cann_context {
      * @brief Destructor for cleaning up resources.
      */
     ~ggml_backend_cann_context() {
+        ggml_cann_set_device(device);
         if (copy_event != nullptr) {
             ACL_CHECK(aclrtDestroyEvent(copy_event));
         }

ggml/src/ggml-cuda.cu

@@ -21,6 +21,8 @@
 #include "ggml-cuda/mmq.cuh"
 #include "ggml-cuda/mmvq.cuh"
 #include "ggml-cuda/norm.cuh"
+#include "ggml-cuda/opt-step-adamw.cuh"
+#include "ggml-cuda/out-prod.cuh"
 #include "ggml-cuda/pad.cuh"
 #include "ggml-cuda/pool2d.cuh"
 #include "ggml-cuda/quantize.cuh"
@@ -32,6 +34,7 @@
 #include "ggml-cuda/tsembd.cuh"
 #include "ggml-cuda/unary.cuh"
 #include "ggml-cuda/upscale.cuh"
+#include "ggml-cuda/rwkv-wkv.cuh"
 
 #include <algorithm>
 #include <array>
@@ -133,7 +136,7 @@ static cudaError_t ggml_cuda_device_malloc(void ** ptr, size_t size, int device)
     return res;
 #else
 
-#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_USE_MUSA)
+#if !defined(GGML_USE_HIPBLAS)
     cudaError_t err;
     if (getenv("GGML_CUDA_ENABLE_UNIFIED_MEMORY") != nullptr)
     {
@@ -146,7 +149,7 @@ static cudaError_t ggml_cuda_device_malloc(void ** ptr, size_t size, int device)
     return err;
 #else
     return cudaMalloc(ptr, size);
-#endif // !defined(GGML_USE_HIPBLAS) && !defined(GGML_USE_MUSA)
+#endif // !defined(GGML_USE_HIPBLAS)
 
 #endif
 }
@@ -184,7 +187,7 @@ static ggml_cuda_device_info ggml_cuda_init() {
     for (int id = 0; id < info.device_count; ++id) {
         int device_vmm = 0;
 
-#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM) && !defined(GGML_USE_MUSA)
+#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM)
         CUdevice device;
         CU_CHECK(cuDeviceGet(&device, id));
         CU_CHECK(cuDeviceGetAttribute(&device_vmm, CU_DEVICE_ATTRIBUTE_VIRTUAL_MEMORY_MANAGEMENT_SUPPORTED, device));
@@ -196,7 +199,7 @@ static ggml_cuda_device_info ggml_cuda_init() {
             alloc_prop.location.id = id;
             CU_CHECK(cuMemGetAllocationGranularity(&info.devices[id].vmm_granularity, &alloc_prop, CU_MEM_ALLOC_GRANULARITY_RECOMMENDED));
         }
-#endif // !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM) && !defined(GGML_USE_MUSA)
+#endif // !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM)
         info.devices[id].vmm = !!device_vmm;
 
         cudaDeviceProp prop;
@@ -332,7 +335,7 @@ struct ggml_cuda_pool_leg : public ggml_cuda_pool {
 };
 
 // pool with virtual memory
-#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM) && !defined(GGML_USE_MUSA)
+#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM)
 struct ggml_cuda_pool_vmm : public ggml_cuda_pool {
     static const size_t CUDA_POOL_VMM_MAX_SIZE = 1ull << 35; // 32 GB
 
@@ -426,14 +429,14 @@ struct ggml_cuda_pool_vmm : public ggml_cuda_pool {
         GGML_ASSERT(ptr == (void *) (pool_addr + pool_used));
     }
 };
-#endif // !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM) && !defined(GGML_USE_MUSA)
+#endif // !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM)
 
 std::unique_ptr<ggml_cuda_pool> ggml_backend_cuda_context::new_pool_for_device(int device) {
-#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM) && !defined(GGML_USE_MUSA)
+#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM)
     if (ggml_cuda_info().devices[device].vmm) {
         return std::unique_ptr<ggml_cuda_pool>(new ggml_cuda_pool_vmm(device));
     }
-#endif // !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM) && !defined(GGML_USE_MUSA)
+#endif // !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM)
     return std::unique_ptr<ggml_cuda_pool>(new ggml_cuda_pool_leg(device));
 }
 
@@ -493,6 +496,14 @@ GGML_CALL static void ggml_backend_cuda_buffer_init_tensor(ggml_backend_buffer_t
     }
 }
 
+GGML_CALL static void ggml_backend_cuda_buffer_memset_tensor(ggml_backend_buffer_t buffer, ggml_tensor * tensor, uint8_t value, size_t offset, size_t size) {
+    ggml_backend_cuda_buffer_context * ctx = (ggml_backend_cuda_buffer_context *)buffer->context;
+
+    ggml_cuda_set_device(ctx->device);
+    CUDA_CHECK(cudaMemsetAsync((char *)tensor->data + offset, value, size, cudaStreamPerThread));
+    CUDA_CHECK(cudaStreamSynchronize(cudaStreamPerThread));
+}
+
 GGML_CALL static void ggml_backend_cuda_buffer_set_tensor(ggml_backend_buffer_t buffer, ggml_tensor * tensor, const void * data, size_t offset, size_t size) {
     ggml_backend_cuda_buffer_context * ctx = (ggml_backend_cuda_buffer_context *)buffer->context;
 
@@ -544,6 +555,7 @@ static ggml_backend_buffer_i ggml_backend_cuda_buffer_interface = {
     /* .free_buffer     = */ ggml_backend_cuda_buffer_free_buffer,
     /* .get_base        = */ ggml_backend_cuda_buffer_get_base,
     /* .init_tensor     = */ ggml_backend_cuda_buffer_init_tensor,
+    /* .memset_tensor   = */ ggml_backend_cuda_buffer_memset_tensor,
     /* .set_tensor      = */ ggml_backend_cuda_buffer_set_tensor,
     /* .get_tensor      = */ ggml_backend_cuda_buffer_get_tensor,
     /* .cpy_tensor      = */ ggml_backend_cuda_buffer_cpy_tensor,
@@ -860,6 +872,7 @@ static struct ggml_backend_buffer_i ggml_backend_cuda_split_buffer_interface = {
     /* .free_buffer     = */ ggml_backend_cuda_split_buffer_free_buffer,
     /* .get_base        = */ ggml_backend_cuda_split_buffer_get_base,
     /* .init_tensor     = */ ggml_backend_cuda_split_buffer_init_tensor,
+    /* .memset_tensor   = */ NULL,
     /* .set_tensor      = */ ggml_backend_cuda_split_buffer_set_tensor,
     /* .get_tensor      = */ ggml_backend_cuda_split_buffer_get_tensor,
     /* .cpy_tensor      = */ NULL,
@@ -2168,6 +2181,9 @@ static bool ggml_cuda_compute_forward(ggml_backend_cuda_context & ctx, struct gg
         case GGML_OP_REPEAT:
             ggml_cuda_op_repeat(ctx, dst);
             break;
+        case GGML_OP_REPEAT_BACK:
+            ggml_cuda_op_repeat_back(ctx, dst);
+            break;
         case GGML_OP_GET_ROWS:
             ggml_cuda_op_get_rows(ctx, dst);
             break;
@@ -2201,6 +2217,9 @@ static bool ggml_cuda_compute_forward(ggml_backend_cuda_context & ctx, struct gg
                 case GGML_UNARY_OP_NEG:
                     ggml_cuda_op_neg(ctx, dst);
                     break;
+                case GGML_UNARY_OP_STEP:
+                    ggml_cuda_op_step(ctx, dst);
+                    break;
                 case GGML_UNARY_OP_GELU:
                     ggml_cuda_op_gelu(ctx, dst);
                     break;
@@ -2225,6 +2244,9 @@ static bool ggml_cuda_compute_forward(ggml_backend_cuda_context & ctx, struct gg
                 case GGML_UNARY_OP_HARDSWISH:
                     ggml_cuda_op_hardswish(ctx, dst);
                     break;
+                case GGML_UNARY_OP_EXP:
+                    ggml_cuda_op_exp(ctx, dst);
+                    break;
                 default:
                     return false;
             }
@@ -2267,6 +2289,9 @@ static bool ggml_cuda_compute_forward(ggml_backend_cuda_context & ctx, struct gg
         case GGML_OP_MUL_MAT_ID:
             ggml_cuda_mul_mat_id(ctx, dst);
             break;
+        case GGML_OP_OUT_PROD:
+            ggml_cuda_out_prod(ctx, dst);
+            break;
         case GGML_OP_SCALE:
             ggml_cuda_op_scale(ctx, dst);
             break;
@@ -2324,6 +2349,15 @@ static bool ggml_cuda_compute_forward(ggml_backend_cuda_context & ctx, struct gg
         case GGML_OP_CROSS_ENTROPY_LOSS:
             ggml_cuda_cross_entropy_loss(ctx, dst);
             break;
+        case GGML_OP_RWKV_WKV:
+            ggml_cuda_op_rwkv_wkv(ctx, dst);
+            break;
+        case GGML_OP_CROSS_ENTROPY_LOSS_BACK:
+            ggml_cuda_cross_entropy_loss_back(ctx, dst);
+            break;
+        case GGML_OP_OPT_STEP_ADAMW:
+            ggml_cuda_opt_step_adamw(ctx, dst);
+            break;
         default:
             return false;
     }
@@ -2451,6 +2485,7 @@ static void set_ggml_graph_node_properties(ggml_tensor * node, ggml_graph_node_p
     for (int i = 0; i < GGML_MAX_SRC; i++) {
         graph_node_properties->src_address[i] = node->src[i] ? node->src[i]->data : nullptr;
     }
+    memcpy(graph_node_properties->op_params, node->op_params, GGML_MAX_OP_PARAMS);
 }
 
 static bool ggml_graph_node_has_matching_properties(ggml_tensor * node, ggml_graph_node_properties * graph_node_properties) {
@@ -2482,6 +2517,12 @@ static bool ggml_graph_node_has_matching_properties(ggml_tensor * node, ggml_gra
             return false;
         }
     }
+
+    if (node->op == GGML_OP_SCALE &&
+        memcmp(graph_node_properties->op_params, node->op_params, GGML_MAX_OP_PARAMS) != 0) {
+        return false;
+    }
+
     return true;
 }
 
@@ -2693,7 +2734,9 @@ GGML_CALL static enum ggml_status ggml_backend_cuda_graph_compute(ggml_backend_t
             // First call with null argument gets number of nodes in graph
             CUDA_CHECK(cudaGraphGetNodes(cuda_ctx->cuda_graph->graph, nullptr, &cuda_ctx->cuda_graph->num_nodes));
             // Subsequent call with non-null argument gets nodes
+            cuda_ctx->cuda_graph->nodes.clear();
             cuda_ctx->cuda_graph->nodes.resize(cuda_ctx->cuda_graph->num_nodes);
+            cuda_ctx->cuda_graph->params.clear();
             cuda_ctx->cuda_graph->params.resize(cuda_ctx->cuda_graph->num_nodes);
             if (cuda_ctx->cuda_graph->num_nodes > 0) {
                 CUDA_CHECK(cudaGraphGetNodes(cuda_ctx->cuda_graph->graph, cuda_ctx->cuda_graph->nodes.data(), &cuda_ctx->cuda_graph->num_nodes));
@@ -2761,6 +2804,7 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
         case GGML_OP_UNARY:
             switch (ggml_get_unary_op(op)) {
                 case GGML_UNARY_OP_NEG:
+                case GGML_UNARY_OP_STEP:
                 case GGML_UNARY_OP_GELU:
                 case GGML_UNARY_OP_SILU:
                 case GGML_UNARY_OP_RELU:
@@ -2769,6 +2813,7 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
                 case GGML_UNARY_OP_HARDSWISH:
                 case GGML_UNARY_OP_GELU_QUICK:
                 case GGML_UNARY_OP_TANH:
+                case GGML_UNARY_OP_EXP:
                     return ggml_is_contiguous(op->src[0]);
                 default:
                     return false;
@@ -2785,6 +2830,12 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
                 if (op->op == GGML_OP_MUL_MAT && a->ne[3] != b->ne[3]) {
                     return false;
                 }
+#ifdef GGML_USE_MUSA
+                if (b->type == GGML_TYPE_F16 && b->ne[2]*b->ne[3] > 1 &&
+                    !ggml_is_transposed(a) && !ggml_is_transposed(b)) {
+                    return false;
+                }
+#endif // GGML_USE_MUSA
                 switch (a->type) {
                     case GGML_TYPE_F32:
                     case GGML_TYPE_F16:
@@ -2808,11 +2859,18 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
                     case GGML_TYPE_IQ3_XXS:
                     case GGML_TYPE_IQ4_NL:
                     case GGML_TYPE_IQ4_XS:
+#ifdef GGML_USE_MUSA
+                        if (a->type == GGML_TYPE_Q3_K) {
+                            return false;
+                        }
+#endif // GGML_USE_MUSA
                         return true;
                     default:
                         return false;
                 }
             } break;
+        case GGML_OP_OUT_PROD:
+            return op->type == GGML_TYPE_F32 && op->src[0]->type == GGML_TYPE_F32 && op->src[1]->type == GGML_TYPE_F32 && op->ne[2] == 1 && op->ne[3] == 1;
         case GGML_OP_GET_ROWS:
             {
                 switch (op->src[0]->type) {
@@ -2841,6 +2899,9 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
                 if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_Q8_0) {
                     return true;
                 }
+                if (src0_type == GGML_TYPE_Q8_0 && src1_type == GGML_TYPE_F32) {
+                    return true;
+                }
                 if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_Q4_0) {
                     return true;
                 }
@@ -2869,6 +2930,12 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
             } break;
         case GGML_OP_DUP:
         case GGML_OP_REPEAT:
+            {
+                ggml_type src0_type = op->src[0]->type;
+                return src0_type != GGML_TYPE_I32 && src0_type != GGML_TYPE_I16;
+            } break;
+        case GGML_OP_REPEAT_BACK:
+                return op->type == GGML_TYPE_F32 && op->src[0]->ne[3] == 1;
         case GGML_OP_CONCAT:
             {
                 ggml_type src0_type = op->src[0]->type;
@@ -2922,22 +2989,28 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
         case GGML_OP_ARANGE:
         case GGML_OP_TIMESTEP_EMBEDDING:
         case GGML_OP_LEAKY_RELU:
+        case GGML_OP_RWKV_WKV:
             return true;
-        case GGML_OP_FLASH_ATTN_EXT:
+        case GGML_OP_FLASH_ATTN_EXT: {
-#if defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)
+#ifndef FLASH_ATTN_AVAILABLE
-            return (op->src[0]->ne[0] == 64 && op->src[1]->type == GGML_TYPE_F16) || op->src[0]->ne[0] == 128;
+            return false;
-#else
+#endif
-            if (op->src[0]->ne[0] == 128) {
-                return true;
-            }
             if (op->src[0]->ne[0] ==  64 && op->src[1]->type == GGML_TYPE_F16) {
                 return true;
             }
-            return ggml_cuda_info().devices[cuda_ctx->device].cc >= CC_VOLTA &&
+            if (op->src[0]->ne[0] == 128) {
-                op->src[1]->type == GGML_TYPE_F16 && op->src[2]->type == GGML_TYPE_F16;
+                return true;
-        case GGML_OP_CROSS_ENTROPY_LOSS:
+            }
+            if (op->src[0]->ne[0] == 256 && op->src[1]->type == GGML_TYPE_F16 && op->src[2]->type == GGML_TYPE_F16) {
+                return true;
+            }
+            const int cc = ggml_cuda_info().devices[cuda_ctx->device].cc;
+            return cc >= CC_VOLTA && cc < CC_OFFSET_AMD && op->src[1]->type == GGML_TYPE_F16 && op->src[2]->type == GGML_TYPE_F16;
+        }
+        case GGML_OP_CROSS_ENTROPY_LOSS:
+        case GGML_OP_CROSS_ENTROPY_LOSS_BACK:
+        case GGML_OP_OPT_STEP_ADAMW:
             return true;
-#endif // defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)
         default:
             return false;
     }

ggml/src/ggml-cuda/binbcast.cu

@@ -1,4 +1,5 @@
 #include "binbcast.cuh"
+#include <cstdint>
 
 static __device__ __forceinline__ float op_repeat(const float a, const float b) {
     return b;
@@ -90,6 +91,30 @@ static __global__ void k_bin_bcast_unravel(const src0_t * src0, const src1_t * s
     dst_row[i0] = (dst_t)bin_op(src0 ? (float)src0_row[i0] : 0.0f, (float)src1_row[i10]);
 }
 
+template <typename T>
+static __global__ void k_repeat_back(
+    const T * __restrict__ src, T * __restrict__ dst, const int64_t ne00, const int64_t ne01, const int64_t ne02,
+    const int64_t ne0, const int64_t ne1, const int64_t ne2) {
+
+    const int64_t tid0 = (int64_t) blockIdx.x*blockDim.x + threadIdx.x;
+    const int64_t tid1 = (int64_t) blockIdx.y*blockDim.y + threadIdx.y;
+    const int64_t tid2 = (int64_t) blockIdx.z*blockDim.z + threadIdx.z;
+
+    if (tid0 >= ne0) {
+        return;
+    }
+
+    T sum = 0;
+    for (int64_t i2 = tid2; i2 < ne02; i2 += ne2) {
+        for (int64_t i1 = tid1; i1 < ne01; i1 += ne1) {
+            for (int64_t i0 = tid0; i0 < ne00; i0 += ne0) {
+                sum += src[i2*ne01*ne00 + i1*ne00 + i0];
+            }
+        }
+    }
+    dst[tid2*ne1*ne0 + tid1*ne0 + tid0] = sum;
+}
+
 template<float (*bin_op)(const float, const float)>
 struct bin_bcast_cuda {
     template<typename src0_t, typename src1_t, typename dst_t>
@@ -247,6 +272,16 @@ struct bin_bcast_cuda {
     }
 };
 
+template <typename T>
+static void repeat_back_cuda(
+    const T * src, T * dst, const int64_t ne00, const int64_t ne01, const int64_t ne02,
+    const int64_t ne0, const int64_t ne1, const int64_t ne2, cudaStream_t stream) {
+
+    const dim3 block_dims(WARP_SIZE, 1, 1);
+    const dim3 block_nums((ne0 + WARP_SIZE - 1) / WARP_SIZE, ne1, ne2);
+    k_repeat_back<T><<<block_nums, block_dims, 0, stream>>>(src, dst, ne00, ne01, ne02, ne0, ne1, ne2);
+}
+
 template<class op>
 static void ggml_cuda_op_bin_bcast(
     const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst,
@@ -286,3 +321,35 @@ void ggml_cuda_op_mul(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
 void ggml_cuda_op_div(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
     ggml_cuda_op_bin_bcast<bin_bcast_cuda<op_div>>(dst->src[0], dst->src[1], dst, dst->src[0]->data, dst->src[1]->data, dst->data, ctx.stream());
 }
+
+void ggml_cuda_op_repeat_back(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
+    const ggml_tensor * src0 = dst->src[0];
+
+    GGML_ASSERT(src0->type == dst->type);
+    GGML_ASSERT(ggml_is_contiguous(src0));
+    GGML_ASSERT(ggml_is_contiguous(dst));
+    GGML_ASSERT(ggml_can_repeat(dst, src0));
+
+    cudaStream_t stream = ctx.stream();
+
+    const int64_t ne00 = src0->ne[0];
+    const int64_t ne01 = src0->ne[1];
+    const int64_t ne02 = src0->ne[2];
+    GGML_ASSERT(src0->ne[3] == 1);
+
+    const int64_t ne0 = dst->ne[0];
+    const int64_t ne1 = dst->ne[1];
+    const int64_t ne2 = dst->ne[2];
+    GGML_ASSERT(dst->ne[3] == 1);
+
+    switch (dst->type) {
+        case GGML_TYPE_F32: {
+            const float * src0_d = (const float *) src0->data;
+            float       * dst_d  = (float       *) dst->data;
+            repeat_back_cuda<float>(src0_d, dst_d, ne00, ne01, ne02, ne0, ne1, ne2, stream);
+        } break;
+        default: {
+            GGML_ASSERT(false);
+        } break;
+    }
+}

ggml/src/ggml-cuda/binbcast.cuh

@@ -5,3 +5,5 @@ void ggml_cuda_op_add(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
 void ggml_cuda_op_sub(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
 void ggml_cuda_op_mul(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
 void ggml_cuda_op_div(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
+
+void ggml_cuda_op_repeat_back(ggml_backend_cuda_context & ctx, ggml_tensor * dst);

ggml/src/ggml-cuda/common.cuh

@@ -50,6 +50,8 @@
 #define CC_RDNA1      (CC_OFFSET_AMD + 1010)
 #define CC_RDNA2      (CC_OFFSET_AMD + 1030)
 #define CC_RDNA3      (CC_OFFSET_AMD + 1100)
+#define CC_QY1        210
+#define CC_QY2        220
 
 #define MATRIX_ROW_PADDING 512 // last row of quant. matrices is a multiple of this to avoid out-of-bounds memory accesses
 
@@ -134,6 +136,10 @@ typedef float2 dfloat2;
 #define INT8_MMA_AVAILABLE
 #endif // !(defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)) && __CUDA_ARCH__ >= CC_TURING
 
+#if !(defined(GGML_USE_MUSA) && __MUSA_ARCH__ <= CC_QY1)
+#define FLASH_ATTN_AVAILABLE
+#endif // !(defined(GGML_USE_MUSA) && __MUSA_ARCH__ <= CC_QY1)
+
 static constexpr bool fast_fp16_available(const int cc) {
     return cc >= CC_PASCAL && cc != 610;
 }
@@ -569,6 +575,7 @@ struct ggml_graph_node_properties {
     int64_t ne[GGML_MAX_DIMS];
     size_t nb[GGML_MAX_DIMS];
     void * src_address[GGML_MAX_SRC];
+    int32_t op_params[GGML_MAX_OP_PARAMS / sizeof(int32_t)];
 };
 
 struct ggml_cuda_graph {

ggml/src/ggml-cuda/cpy.cu

@@ -81,6 +81,17 @@ static __device__ void cpy_blck_f32_q8_0(const char * cxi, char * cdsti) {
     }
 }
 
+static __device__ void cpy_blck_q8_0_f32(const char * cxi, char * cdsti) {
+    const block_q8_0 * xi = (const block_q8_0 *) cxi;
+    float * dsti = (float *) cdsti;
+
+    const float d = (float)xi->d;
+
+    for (int j = 0; j < QK8_0; j++) {
+       dsti[j] = xi->qs[j] * d;
+    }
+}
+
 static __device__ void cpy_blck_f32_q4_0(const char * cxi, char * cdsti) {
     const float * xi = (const float *) cxi;
     block_q4_0 * dsti = (block_q4_0 *) cdsti;
@@ -288,6 +299,32 @@ static __global__ void cpy_f32_q(const char * cx, char * cdst, const int ne,
     cpy_blck(cx + x_offset, cdst + dst_offset);
 }
 
+template <cpy_kernel_t cpy_blck, int qk>
+static __global__ void cpy_q_f32(const char * cx, char * cdst, const int ne,
+                                 const int ne00, const int ne01, const int ne02, const int nb00, const int nb01, const int nb02,
+                                 const int nb03, const int ne10, const int ne11, const int ne12, const int nb10, const int nb11,
+                                 const int nb12, const int nb13) {
+    const int i = (blockDim.x*blockIdx.x + threadIdx.x)*qk;
+
+    if (i >= ne) {
+        return;
+    }
+
+    const int i03 = i/(ne00 * ne01 * ne02);
+    const int i02 = (i - i03*ne00*ne01*ne02 )/ (ne00*ne01);
+    const int i01 = (i - i03*ne00*ne01*ne02  -  i02*ne01*ne00) / ne00;
+    const int i00 = i - i03*ne00*ne01*ne02 - i02*ne01*ne00 - i01*ne00;
+    const int x_offset = (i00/qk)*nb00 + i01*nb01 + i02*nb02 + i03 * nb03;
+
+    const int i13 = i/(ne10 * ne11 * ne12);
+    const int i12 = (i - i13*ne10*ne11*ne12) / (ne10*ne11);
+    const int i11 = (i - i13*ne10*ne11*ne12 - i12*ne10*ne11) / ne10;
+    const int i10 = i - i13*ne10*ne11*ne12 - i12*ne10*ne11 - i11*ne10;
+    const int dst_offset = i10*nb10 + i11*nb11 + i12*nb12 + i13*nb13;
+
+    cpy_blck(cx + x_offset, cdst + dst_offset);
+}
+
 static void ggml_cpy_f16_f32_cuda(
     const char * cx, char * cdst, const int ne,
     const int ne00, const int ne01, const int ne02, const int nb00, const int nb01, const int nb02,
@@ -329,6 +366,16 @@ static void ggml_cpy_f32_q8_0_cuda(
         (cx, cdst, ne, ne00, ne01, ne02, nb00, nb01, nb02, nb03, ne10, ne11, ne12, nb10, nb11, nb12, nb13);
 }
 
+static void ggml_cpy_q8_0_f32_cuda(
+    const char * cx, char * cdst, const int ne,
+    const int ne00, const int ne01, const int ne02, const int nb00, const int nb01, const int nb02,
+    const int nb03, const int ne10, const int ne11, const int ne12, const int nb10, const int nb11, const int nb12, const int nb13, cudaStream_t stream) {
+
+    const int num_blocks = ne;
+    cpy_q_f32<cpy_blck_q8_0_f32, QK8_0><<<num_blocks, 1, 0, stream>>>
+        (cx, cdst, ne, ne00, ne01, ne02, nb00, nb01, nb02, nb03, ne10, ne11, ne12, nb10, nb11, nb12, nb13);
+}
+
 static void ggml_cpy_f32_q4_0_cuda(
     const char * cx, char * cdst, const int ne,
     const int ne00, const int ne01, const int ne02, const int nb00, const int nb01, const int nb02,
@@ -437,6 +484,8 @@ void ggml_cuda_cpy(ggml_backend_cuda_context & ctx, const ggml_tensor * src0, gg
         ggml_cpy_f32_f16_cuda (src0_ddc, src1_ddc, ne, ne00, ne01, ne02, nb00, nb01, nb02, nb03, ne10, ne11, ne12, nb10, nb11, nb12, nb13, main_stream);
     } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_Q8_0) {
         ggml_cpy_f32_q8_0_cuda(src0_ddc, src1_ddc, ne, ne00, ne01, ne02, nb00, nb01, nb02, nb03, ne10, ne11, ne12, nb10, nb11, nb12, nb13, main_stream);
+    } else if (src0->type == GGML_TYPE_Q8_0 && src1->type == GGML_TYPE_F32) {
+        ggml_cpy_q8_0_f32_cuda(src0_ddc, src1_ddc, ne, ne00, ne01, ne02, nb00, nb01, nb02, nb03, ne10, ne11, ne12, nb10, nb11, nb12, nb13, main_stream);
     } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_Q4_0) {
         ggml_cpy_f32_q4_0_cuda(src0_ddc, src1_ddc, ne, ne00, ne01, ne02, nb00, nb01, nb02, nb03, ne10, ne11, ne12, nb10, nb11, nb12, nb13, main_stream);
     } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_Q4_1) {
@@ -471,6 +520,8 @@ void* ggml_cuda_cpy_fn(const ggml_tensor * src0, ggml_tensor * src1) {
         return (void*) cpy_f32_f16<cpy_1_f32_f16>;
     } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_Q8_0) {
         return (void*) cpy_f32_q<cpy_blck_f32_q8_0, QK8_0>;
+    } else if (src0->type == GGML_TYPE_Q8_0 && src1->type == GGML_TYPE_F32) {
+        return (void*) cpy_q_f32<cpy_blck_q8_0_f32, QK8_0>;
     } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_Q4_0) {
         return (void*) cpy_f32_q<cpy_blck_f32_q4_0, QK4_0>;
     } else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_Q4_1) {

ggml/src/ggml-cuda/cross-entropy-loss.cu

@@ -71,6 +71,32 @@ static __global__ void cross_entropy_loss_f32(const float * logits, const float
     dst[blockIdx.x] = loss;
 }
 
+static __global__ void cross_entropy_loss_back_f32(const float * logits, const float * labels, const float * loss, float * dst, const int nclasses) {
+    extern __shared__ float tmp[];
+
+    float maxval = -INFINITY;
+    for (int i = threadIdx.x; i < nclasses; i += WARP_SIZE) {
+        const float val = logits[blockIdx.x*nclasses + i];
+        maxval = fmaxf(maxval, val);
+        tmp[i] = val;
+    }
+    maxval = warp_reduce_max(maxval);
+
+    float sum = 0.0f;
+    for (int i = threadIdx.x; i < nclasses; i += WARP_SIZE) {
+        const float val = expf(tmp[i] - maxval);
+        sum += val;
+        tmp[i] = val;
+    }
+    sum = warp_reduce_sum(sum);
+    const float sm_scale = 1.0f/sum;
+
+    const float d_by_nrows = *loss/gridDim.x;
+    for (int i = threadIdx.x; i < nclasses; i += WARP_SIZE) {
+        dst[blockIdx.x*nclasses + i] = (tmp[i]*sm_scale - labels[blockIdx.x*nclasses + i])*d_by_nrows;
+    }
+}
+
 void ggml_cuda_cross_entropy_loss(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
     const ggml_tensor * src0 = dst->src[0];
     const ggml_tensor * src1 = dst->src[1];
@@ -104,3 +130,37 @@ void ggml_cuda_cross_entropy_loss(ggml_backend_cuda_context & ctx, ggml_tensor *
     // Combine results from individual blocks:
     sum_f32_cuda(pool, dst_tmp.ptr, dst_d, blocks_num.x, stream);
 }
+
+void ggml_cuda_cross_entropy_loss_back(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
+    const ggml_tensor * src0 = dst->src[0];
+    const ggml_tensor * src1 = dst->src[1];
+    const ggml_tensor * opt0 = dst->src[2];
+
+    GGML_ASSERT(src0->type == GGML_TYPE_F32);
+    GGML_ASSERT(src1->type == GGML_TYPE_F32);
+    GGML_ASSERT(opt0->type == GGML_TYPE_F32);
+    GGML_ASSERT( dst->type == GGML_TYPE_F32);
+
+    GGML_ASSERT(ggml_is_contiguous(src0));
+    GGML_ASSERT(ggml_is_contiguous(src1));
+    GGML_ASSERT(ggml_is_contiguous(opt0));
+    GGML_ASSERT(ggml_is_contiguous(dst));
+    GGML_ASSERT(ggml_are_same_shape(src0, src1));
+    GGML_ASSERT(ggml_are_same_shape(src0, dst));
+
+    const int64_t ne00  = src0->ne[0];
+    const int64_t nrows = ggml_nrows(src0);
+
+    const float * src0_d = (const float *) src0->data;
+    const float * src1_d = (const float *) src1->data;
+    const float * opt0_d = (const float *) opt0->data;
+    float       * dst_d  = (float       *) dst->data;
+
+    cudaStream_t stream = ctx.stream();
+
+    const dim3 blocks_dim(WARP_SIZE, 1, 1);
+    const dim3 blocks_num(nrows, 1, 1);
+    const int shmem = ne00*sizeof(float);
+
+    cross_entropy_loss_back_f32<<<blocks_num, blocks_dim, shmem, stream>>>(src0_d, src1_d, opt0_d, dst_d, ne00);
+}

ggml/src/ggml-cuda/cross-entropy-loss.cuh

@@ -3,3 +3,5 @@
 #define CUDA_CROSS_ENTROPY_LOSS_BLOCK_SIZE 256
 
 void ggml_cuda_cross_entropy_loss(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
+
+void ggml_cuda_cross_entropy_loss_back(ggml_backend_cuda_context & ctx, ggml_tensor * dst);

ggml/src/ggml-cuda/fattn-tile-f32.cu

@@ -44,13 +44,17 @@ static __global__ void flash_attn_tile_ext_f32(
         const int ne1,
         const int ne2,
         const int ne3) {
+#ifndef FLASH_ATTN_AVAILABLE
+    NO_DEVICE_CODE;
+    return;
+#endif // FLASH_ATTN_AVAILABLE
     // Skip unused kernel variants for faster compilation:
     if (use_logit_softcap && !(D == 128 || D == 256)) {
         NO_DEVICE_CODE;
         return;
     }
 
-    //In this kernel Q, K, V are matrices while i, j, k are matrix indices.
+    // In this kernel Q, K, V are matrices while i, j, k are matrix indices.
 
     const int ic0 = (blockIdx.x / parallel_blocks) * ncols; // Index of the Q/QKV column to work on.
     const int ip  =  blockIdx.x % parallel_blocks; // Index in group of blocks running for the same column in parallel.

ggml/src/ggml-cuda/fattn.cu

@@ -314,7 +314,7 @@ void ggml_cuda_flash_attn_ext(ggml_backend_cuda_context & ctx, ggml_tensor * dst
     }
 
     if (!fast_fp16_available(cc)) {
-        if (Q->ne[1] <= 8) {
+        if (Q->ne[1] <= 8 || Q->ne[0] == 256) {
             ggml_cuda_flash_attn_ext_vec_f32(ctx, dst);
         } else {
             ggml_cuda_flash_attn_ext_tile_f32(ctx, dst);

ggml/src/ggml-cuda/im2col.cu

@@ -69,7 +69,6 @@ void ggml_cuda_op_im2col(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
     float * dst_d = (float *)dst->data;
     cudaStream_t stream = ctx.stream();
 
-    GGML_ASSERT(src0->type == GGML_TYPE_F16);
     GGML_ASSERT(src1->type == GGML_TYPE_F32);
     GGML_ASSERT( dst->type == GGML_TYPE_F16 || dst->type == GGML_TYPE_F32);
 

ggml/src/ggml-cuda/opt-step-adamw.cu

@@ -0,0 +1,80 @@
+#include "opt-step-adamw.cuh"
+
+#include <cstdint>
+
+static __global__ void opt_step_adamw_f32(
+    float * __restrict__ x, const float * __restrict__ g, float * __restrict__ g_m, float * __restrict__ g_v, const int64_t k,
+    const float alpha, const float beta1, const float beta2, const float eps, const float wd,
+    const float beta1h, const float beta2h) {
+
+    const int64_t i = (int64_t) blockIdx.x*blockDim.x + threadIdx.x;
+
+    if (i >= k) {
+        return;
+    }
+
+    const float gi = g[i];
+    const float gmi = g_m[i]*beta1 +    gi*(1.0f - beta1);
+    const float gvi = g_v[i]*beta2 + gi*gi*(1.0f - beta2);
+
+    g_m[i] = gmi;
+    g_v[i] = gvi;
+
+    const float mh =       gmi*beta1h;
+    const float vh = sqrtf(gvi*beta2h) + eps;
+
+    x[i] = x[i]*(1.0f - alpha*wd) - mh/vh;
+}
+
+static void opt_step_adamw_f32_cuda(
+    float * x, const float * g, float * g_m, float * g_v, const int64_t k,
+    const float alpha, const float beta1, const float beta2, const float eps, const float wd,
+    const float beta1h, const float beta2h, cudaStream_t stream) {
+
+    const dim3 block_dims(CUDA_OPT_STEP_ADAMW_BLOCK_SIZE, 1, 1);
+    const dim3 block_nums((k + CUDA_OPT_STEP_ADAMW_BLOCK_SIZE - 1) / CUDA_OPT_STEP_ADAMW_BLOCK_SIZE, 1, 1);
+    opt_step_adamw_f32<<<block_nums, block_dims, 0, stream>>>(x, g, g_m, g_v, k, alpha, beta1, beta2, eps, wd, beta1h, beta2h);
+}
+
+void ggml_cuda_opt_step_adamw(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
+    const ggml_tensor * src0        = dst->src[0];
+    const ggml_tensor * src0_grad   = dst->src[1];
+    const ggml_tensor * src0_grad_m = dst->src[2];
+    const ggml_tensor * src0_grad_v = dst->src[3];
+
+    GGML_ASSERT(src0->type        == GGML_TYPE_F32);
+    GGML_ASSERT(src0_grad->type   == GGML_TYPE_F32);
+    GGML_ASSERT(src0_grad_m->type == GGML_TYPE_F32);
+    GGML_ASSERT(src0_grad_v->type == GGML_TYPE_F32);
+    GGML_ASSERT(ggml_is_contiguous(src0));
+    GGML_ASSERT(ggml_is_contiguous(src0_grad));
+    GGML_ASSERT(ggml_is_contiguous(src0_grad_m));
+    GGML_ASSERT(ggml_is_contiguous(src0_grad_v));
+    GGML_ASSERT(ggml_are_same_shape(src0, src0_grad));
+    GGML_ASSERT(ggml_are_same_shape(src0, src0_grad_m));
+    GGML_ASSERT(ggml_are_same_shape(src0, src0_grad_v));
+
+    float       * src0_d        = (float       *) src0->data;
+    const float * src0_grad_d   = (const float *) src0_grad->data;
+    float       * src0_grad_m_d = (float       *) src0_grad_m->data;
+    float       * src0_grad_v_d = (float       *) src0_grad_v->data;
+
+    cudaStream_t stream = ctx.stream();
+
+    const int64_t ne = ggml_nelements(src0);
+
+    int64_t iter;  memcpy(&iter,  &dst->op_params[0], sizeof(int64_t));
+    float   alpha; memcpy(&alpha, &dst->op_params[2], sizeof(float));
+    float   beta1; memcpy(&beta1, &dst->op_params[3], sizeof(float));
+    float   beta2; memcpy(&beta2, &dst->op_params[4], sizeof(float));
+    float   eps;   memcpy(&eps,   &dst->op_params[5], sizeof(float));
+    float   wd;    memcpy(&wd,    &dst->op_params[6], sizeof(float));
+
+    const float beta1h  = alpha/(1.0f - powf(beta1, iter));
+    const float beta2h  =  1.0f/(1.0f - powf(beta2, iter));
+
+    opt_step_adamw_f32_cuda(src0_d, src0_grad_d, src0_grad_m_d, src0_grad_v_d, ne, alpha, beta1, beta2, eps, wd, beta1h, beta2h, stream);
+
+    iter++;
+    memcpy(&dst->op_params[0], &iter, sizeof(int64_t));
+}

ggml/src/ggml-cuda/opt-step-adamw.cuh

@@ -0,0 +1,5 @@
+#include "common.cuh"
+
+#define CUDA_OPT_STEP_ADAMW_BLOCK_SIZE 256
+
+void ggml_cuda_opt_step_adamw(ggml_backend_cuda_context & ctx, ggml_tensor * dst);

ggml/src/ggml-cuda/out-prod.cu

@@ -0,0 +1,51 @@
+#include "out-prod.cuh"
+
+#include <cstdint>
+
+void ggml_cuda_out_prod(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
+    const ggml_tensor * src0 = dst->src[0];
+    const ggml_tensor * src1 = dst->src[1];
+
+    GGML_TENSOR_BINARY_OP_LOCALS
+
+    GGML_ASSERT(src0->type == GGML_TYPE_F32);
+    GGML_ASSERT(src1->type == GGML_TYPE_F32);
+    GGML_ASSERT(dst->type  == GGML_TYPE_F32);
+    GGML_ASSERT(ggml_is_contiguous(src0));
+    GGML_ASSERT(ggml_is_contiguous(dst));
+
+    GGML_ASSERT(ne01 == ne11);
+    GGML_ASSERT(ne0 == ne00);
+    GGML_ASSERT(ne1 == ne10);
+
+    GGML_ASSERT(ne2 == src0->ne[2]);
+    GGML_ASSERT(ne2 == src1->ne[2]);
+    GGML_ASSERT(ne3 == src0->ne[3]);
+    GGML_ASSERT(ne3 == src1->ne[3]);
+
+    const float * src0_d = (const float *) src0->data;
+    const float * src1_d = (const float *) src1->data;
+    float       *  dst_d = (float       *)  dst->data;
+
+    cudaStream_t   stream = ctx.stream();
+    cublasHandle_t handle = ctx.cublas_handle();
+
+    const float alpha = 1.0f;
+    const float beta = 0.0f;
+
+    GGML_ASSERT(ne2 == 1);
+    GGML_ASSERT(ne3 == 1);
+    CUBLAS_CHECK(cublasSetStream(handle, stream));
+
+    const bool src1_T = ggml_is_transposed(src1);
+    const cublasOperation_t src1_cublas_op =  src1_T ? CUBLAS_OP_N : CUBLAS_OP_T;
+    const int64_t           ldb            = (src1_T ?        nb10 :        nb11) /  sizeof(float);
+    GGML_ASSERT(                             (src1_T ?        nb11 :        nb10) == sizeof(float));
+
+    CUBLAS_CHECK(
+        cublasSgemm(handle, CUBLAS_OP_N, src1_cublas_op,
+                ne0, ne1, ne01,
+                &alpha, src0_d, ne00,
+                        src1_d, ldb,
+                &beta,  dst_d,  ne0));
+}

ggml/src/ggml-cuda/out-prod.cuh

@@ -0,0 +1,3 @@
+#include "common.cuh"
+
+void ggml_cuda_out_prod(ggml_backend_cuda_context & ctx, ggml_tensor * dst);

ggml/src/ggml-cuda/rwkv-wkv.cu

@@ -0,0 +1,89 @@
+#include "common.cuh"
+#include "rwkv-wkv.cuh"
+
+static __global__ void rwkv_wkv_f32(const int B, const int T, const int C, const int H, const float * k, const float * v, const float * r, const float * tf, const float * td, const float * s, float * dst) {
+    const int tid = threadIdx.x;
+    const int bid = blockIdx.x;
+
+    const int head_size = CUDA_WKV_BLOCK_SIZE;
+    const int batch_i = bid / H;
+    const int head_i = bid % H;
+    const int state_size = C * head_size;
+    const int n_seq_tokens = T / B;
+
+    float state[head_size];
+    __shared__ float _k[head_size], _r[head_size], _tf[head_size], _td[head_size];
+
+    #pragma unroll
+    for (int i = 0; i < head_size; i++) {
+        state[i] = s[batch_i * state_size + head_i * head_size * head_size + i * head_size + tid];
+    }
+
+    __syncthreads();
+    _tf[tid] = tf[head_i * head_size + tid];
+    __syncthreads();
+
+    for (int t = batch_i * n_seq_tokens * C + head_i * head_size + tid; t < (batch_i + 1) * n_seq_tokens * C + head_i * head_size + tid; t += C) {
+        __syncthreads();
+        _k[tid] = k[t];
+        _r[tid] = r[t];
+        _td[tid] = td[t];
+        __syncthreads();
+
+        const float _v = v[t];
+        float y = 0;
+        for (int j = 0; j < head_size; j += 4) {
+            const float4& k = (float4&)(_k[j]);
+            const float4& r = (float4&)(_r[j]);
+            const float4& tf = (float4&)(_tf[j]);
+            const float4& td = (float4&)(_td[j]);
+            float4& s = (float4&)(state[j]);
+            float4 kv;
+
+            kv.x = k.x * _v;
+            kv.y = k.y * _v;
+            kv.z = k.z * _v;
+            kv.w = k.w * _v;
+
+            y += r.x * (tf.x * kv.x + s.x);
+            y += r.y * (tf.y * kv.y + s.y);
+            y += r.z * (tf.z * kv.z + s.z);
+            y += r.w * (tf.w * kv.w + s.w);
+
+            s.x = s.x * td.x + kv.x;
+            s.y = s.y * td.y + kv.y;
+            s.z = s.z * td.z + kv.z;
+            s.w = s.w * td.w + kv.w;
+        }
+        dst[t] = y;
+    }
+
+    #pragma unroll
+    for (int i = 0; i < head_size; i++) {
+        dst[T * C + batch_i * state_size + head_i * head_size * head_size + i * head_size + tid] = state[i];
+    }
+}
+
+void ggml_cuda_op_rwkv_wkv(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
+    const float * k_d  = (const float *)dst->src[0]->data;
+    const float * v_d  = (const float *)dst->src[1]->data;
+    const float * r_d  = (const float *)dst->src[2]->data;
+    const float * tf_d = (const float *)dst->src[3]->data;
+    const float * td_d = (const float *)dst->src[4]->data;
+    const float * s_d  = (const float *)dst->src[5]->data;
+
+    const int64_t B = dst->src[5]->ne[1];
+    const int64_t T = dst->src[0]->ne[3];
+    const int64_t C = dst->ne[0];
+    const int64_t H = dst->src[0]->ne[2];
+
+    float * dst_d = (float *)dst->data;
+
+    cudaStream_t stream = ctx.stream();
+
+    GGML_ASSERT(dst->src[5]->type == GGML_TYPE_F32);
+    GGML_ASSERT(C % H == 0);
+    GGML_ASSERT(C / H == CUDA_WKV_BLOCK_SIZE);
+
+    rwkv_wkv_f32<<<B * H, C / H, 0, stream>>>(B, T, C, H, k_d, v_d, r_d, tf_d, td_d, s_d, dst_d);
+}

ggml/src/ggml-cuda/rwkv-wkv.cuh

@@ -0,0 +1,5 @@
+#include "common.cuh"
+
+#define CUDA_WKV_BLOCK_SIZE 64
+
+void ggml_cuda_op_rwkv_wkv(ggml_backend_cuda_context & ctx, ggml_tensor * dst);

ggml/src/ggml-cuda/sum.cu

@@ -1,9 +1,13 @@
-#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_USE_MUSA)
+#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_USE_MUSA) && CUDART_VERSION >= 11700
+#define USE_CUB
+#endif // !defined(GGML_USE_HIPBLAS) && !defined(GGML_USE_MUSA) && CUDART_VERSION >= 11700
+
+#ifdef USE_CUB
 // On Windows CUB uses libraries with variables called CC_PASCAL which conflict with the define in common.cuh.
 // For this reason CUB must be included BEFORE anything else.
 #include <cub/cub.cuh>
 using namespace cub;
-#endif // !defined(GGML_USE_HIPBLAS) && !defined(GGML_USE_MUSA)
+#endif // USE_CUB
 
 #include "sumrows.cuh"
 #include "sum.cuh"
@@ -11,7 +15,7 @@ using namespace cub;
 #include <cstdint>
 
 void sum_f32_cuda(ggml_cuda_pool & pool, const float * x, float * dst, const int64_t ne, cudaStream_t stream) {
-#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_USE_MUSA)
+#ifdef USE_CUB
     size_t tmp_size = 0;
     DeviceReduce::Sum(nullptr,       tmp_size, x, dst, ne, stream);
     ggml_cuda_pool_alloc<uint8_t> tmp_alloc(pool, tmp_size);
@@ -21,7 +25,7 @@ void sum_f32_cuda(ggml_cuda_pool & pool, const float * x, float * dst, const int
     // For AMD there is rocPRIM which could be used as a drop-in replacement via hipcub but this would require C++11 -> C++14.
     sum_rows_f32_cuda(x, dst, ne, 1, stream);
     GGML_UNUSED(pool);
-#endif // !defined(GGML_USE_HIPBLAS) && !defined(GGML_USE_MUSA)
+#endif // USE_CUB
 }
 
 void ggml_cuda_op_sum(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {

ggml/src/ggml-cuda/unary.cu

@@ -10,6 +10,16 @@ static __global__ void neg_f32(const float * x, float * dst, const int k) {
     dst[i] = -x[i];
 }
 
+static __global__ void step_f32(const float * x, float * dst, const int k) {
+    const int i = blockDim.x*blockIdx.x + threadIdx.x;
+
+    if (i >= k) {
+        return;
+    }
+
+    dst[i] = x[i] > 0.0f;
+}
+
 static __global__ void gelu_f32(const float * x, float * dst, const int k) {
     const float GELU_COEF_A    = 0.044715f;
     const float SQRT_2_OVER_PI = 0.79788456080286535587989211986876f;
@@ -85,6 +95,15 @@ static __global__ void hardswish_f32(const float * x, float * dst, const int k)
     dst[i] = x[i] * fminf(1.0f, fmaxf(0.0f, (x[i] + 3.0f) / 6.0f));
 }
 
+static __global__ void exp_f32(const float * x, float * dst, const int k) {
+    const int i = blockDim.x*blockIdx.x + threadIdx.x;
+
+    if (i >= k) {
+        return;
+    }
+    dst[i] = expf(x[i]);
+}
+
 static __global__ void leaky_relu_f32(const float * x, float * dst, const int k, const float negative_slope) {
     const int i  = blockDim.x*blockIdx.x + threadIdx.x;
     if (i >= k) {
@@ -134,6 +153,11 @@ static void neg_f32_cuda(const float * x, float * dst, const int k, cudaStream_t
     neg_f32<<<num_blocks, CUDA_NEG_BLOCK_SIZE, 0, stream>>>(x, dst, k);
 }
 
+static void step_f32_cuda(const float * x, float * dst, const int k, cudaStream_t stream) {
+    const int num_blocks = (k + CUDA_STEP_BLOCK_SIZE - 1) / CUDA_STEP_BLOCK_SIZE;
+    step_f32<<<num_blocks, CUDA_STEP_BLOCK_SIZE, 0, stream>>>(x, dst, k);
+}
+
 static void gelu_f32_cuda(const float * x, float * dst, const int k, cudaStream_t stream) {
     const int num_blocks = (k + CUDA_GELU_BLOCK_SIZE - 1) / CUDA_GELU_BLOCK_SIZE;
     gelu_f32<<<num_blocks, CUDA_GELU_BLOCK_SIZE, 0, stream>>>(x, dst, k);
@@ -174,6 +198,11 @@ static void hardswish_f32_cuda(const float * x, float * dst, const int k, cudaSt
     hardswish_f32<<<num_blocks, CUDA_HARDSWISH_BLOCK_SIZE, 0, stream>>>(x, dst, k);
 }
 
+static void exp_f32_cuda(const float * x, float * dst, const int k, cudaStream_t stream) {
+    const int num_blocks = (k + CUDA_EXP_BLOCK_SIZE - 1) / CUDA_EXP_BLOCK_SIZE;
+    exp_f32<<<num_blocks, CUDA_EXP_BLOCK_SIZE, 0, stream>>>(x, dst, k);
+}
+
 static void leaky_relu_f32_cuda(const float * x, float * dst, const int k, const float negative_slope, cudaStream_t stream) {
     const int num_blocks = (k + CUDA_RELU_BLOCK_SIZE - 1) / CUDA_RELU_BLOCK_SIZE;
     leaky_relu_f32<<<num_blocks, CUDA_RELU_BLOCK_SIZE, 0, stream>>>(x, dst, k, negative_slope);
@@ -213,6 +242,20 @@ void ggml_cuda_op_neg(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
     neg_f32_cuda(src0_d, dst_d, ggml_nelements(src0), stream);
 }
 
+void ggml_cuda_op_step(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
+    const ggml_tensor * src0 = dst->src[0];
+    const float * src0_d = (const float *)src0->data;
+    float * dst_d = (float *)dst->data;
+    cudaStream_t stream = ctx.stream();
+
+    GGML_ASSERT(ggml_is_contiguous(src0));
+
+    GGML_ASSERT(src0->type == GGML_TYPE_F32);
+    GGML_ASSERT( dst->type == GGML_TYPE_F32);
+
+    step_f32_cuda(src0_d, dst_d, ggml_nelements(src0), stream);
+}
+
 void ggml_cuda_op_gelu(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
     const ggml_tensor * src0 = dst->src[0];
     const float * src0_d = (const float *)src0->data;
@@ -325,6 +368,20 @@ void ggml_cuda_op_hardswish(ggml_backend_cuda_context & ctx, ggml_tensor * dst)
     hardswish_f32_cuda(src0_d, dst_d, ggml_nelements(src0), stream);
 }
 
+void ggml_cuda_op_exp(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
+    const ggml_tensor * src0 = dst->src[0];
+    const float * src0_d = (const float *)src0->data;
+    float * dst_d = (float *)dst->data;
+    cudaStream_t stream = ctx.stream();
+
+    GGML_ASSERT(ggml_is_contiguous(src0));
+
+    GGML_ASSERT(src0->type == GGML_TYPE_F32);
+    GGML_ASSERT( dst->type == GGML_TYPE_F32);
+
+    exp_f32_cuda(src0_d, dst_d, ggml_nelements(src0), stream);
+}
+
 void ggml_cuda_op_leaky_relu(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
     const ggml_tensor * src0 = dst->src[0];
     const float * src0_d = (const float *)src0->data;

ggml/src/ggml-cuda/unary.cuh

@@ -1,12 +1,14 @@
 #include "common.cuh"
 
 #define CUDA_NEG_BLOCK_SIZE 256
+#define CUDA_STEP_BLOCK_SIZE 256
 #define CUDA_GELU_BLOCK_SIZE 256
 #define CUDA_SILU_BLOCK_SIZE 256
 #define CUDA_TANH_BLOCK_SIZE 256
 #define CUDA_RELU_BLOCK_SIZE 256
 #define CUDA_SIGMOID_BLOCK_SIZE 256
 #define CUDA_HARDSIGMOID_BLOCK_SIZE 256
+#define CUDA_EXP_BLOCK_SIZE 256
 #define CUDA_HARDSWISH_BLOCK_SIZE 256
 #define CUDA_SQR_BLOCK_SIZE 256
 #define CUDA_SQRT_BLOCK_SIZE 256
@@ -15,6 +17,8 @@
 
 void ggml_cuda_op_neg(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
 
+void ggml_cuda_op_step(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
+
 void ggml_cuda_op_gelu(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
 
 void ggml_cuda_op_silu(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
@@ -29,6 +33,8 @@ void ggml_cuda_op_sigmoid(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
 
 void ggml_cuda_op_hardsigmoid(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
 
+void ggml_cuda_op_exp(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
+
 void ggml_cuda_op_hardswish(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
 
 void ggml_cuda_op_leaky_relu(ggml_backend_cuda_context & ctx, ggml_tensor * dst);

ggml/src/ggml-cuda/vendors/hip.h

@@ -30,6 +30,7 @@
 #define cublasSetStream hipblasSetStream
 #define cublasSgemm hipblasSgemm
 #define cublasStatus_t hipblasStatus_t
+#define cublasOperation_t hipblasOperation_t
 #define cudaDataType_t hipblasDatatype_t //deprecated, new hipblasDatatype not in 5.6
 #define cudaDeviceCanAccessPeer hipDeviceCanAccessPeer
 #define cudaDeviceDisablePeerAccess hipDeviceDisablePeerAccess

ggml/src/ggml-cuda/vendors/musa.h

@@ -26,6 +26,7 @@
 #define cublasSetStream mublasSetStream
 #define cublasSgemm mublasSgemm
 #define cublasStatus_t mublasStatus_t
+#define cublasOperation_t mublasOperation_t
 #define cublasGetStatusString mublasStatus_to_string
 #define cudaDataType_t musaDataType_t
 #define cudaDeviceCanAccessPeer musaDeviceCanAccessPeer
@@ -56,6 +57,7 @@
 #define cudaLaunchHostFunc musaLaunchHostFunc
 #define cudaMalloc musaMalloc
 #define cudaMallocHost musaMallocHost
+#define cudaMallocManaged musaMallocManaged
 #define cudaMemcpy musaMemcpy
 #define cudaMemcpyAsync musaMemcpyAsync
 #define cudaMemcpyPeerAsync musaMemcpyPeerAsync

ggml/src/ggml-kompute.cpp

@@ -1872,6 +1872,7 @@ static ggml_backend_buffer_i ggml_backend_kompute_buffer_i = {
     /* .free_buffer     = */ ggml_backend_kompute_buffer_free_buffer,
     /* .get_base        = */ ggml_backend_kompute_buffer_get_base,
     /* .init_tensor     = */ NULL,
+    /* .memset_tensor   = */ NULL,
     /* .set_tensor      = */ ggml_backend_kompute_buffer_set_tensor,
     /* .get_tensor      = */ ggml_backend_kompute_buffer_get_tensor,
     /* .cpy_tensor      = */ NULL,

ggml/src/ggml-metal.m

@@ -12,6 +12,12 @@
 #define MIN(a, b) ((a) < (b) ? (a) : (b))
 #define MAX(a, b) ((a) > (b) ? (a) : (b))
 
+// max memory buffers that can be mapped to the device
+#define GGML_METAL_MAX_BUFFERS 64
+
+// max number of MTLCommandBuffer used to submit a graph for processing
+#define GGML_METAL_MAX_COMMAND_BUFFERS 8
+
 #ifdef GGML_METAL_NDEBUG
 #define GGML_METAL_LOG(...)
 #define GGML_METAL_LOG_INFO(...)
@@ -221,11 +227,11 @@ enum ggml_metal_kernel_type {
 };
 
 struct ggml_backend_metal_context {
-    int n_cb;
-
     id<MTLDevice>       device;
     id<MTLCommandQueue> queue;
 
+    MTLComputePassDescriptor * edesc;
+
     dispatch_queue_t d_queue;
 
     struct ggml_metal_kernel kernels[GGML_METAL_KERNEL_TYPE_COUNT];
@@ -233,7 +239,27 @@ struct ggml_backend_metal_context {
     bool support_simdgroup_reduction;
     bool support_simdgroup_mm;
 
-    bool should_capture_next_compute;
+    // capture state
+    bool capture_next_compute;
+    bool capture_started;
+
+    id<MTLCaptureScope> capture_scope;
+
+    // command buffer state
+    int n_cb;           // number of extra threads used to submit the command buffers
+    int n_nodes_0;      // number of nodes submitted by the main thread
+    int n_nodes_1;      // remaining number of nodes submitted by the n_cb threads
+    int n_nodes_per_cb;
+
+    struct ggml_cgraph * gf;
+
+    // the callback given to the thread pool
+    // TODO: ideally, this should be created once, utilizing the command buffer state above
+    //       for some reason, doing it like this leads to a crash
+    void (^encode_async)(size_t ith);
+
+    // n_cb command buffers + 1 used by the main thread
+    id<MTLCommandBuffer> command_buffers[GGML_METAL_MAX_COMMAND_BUFFERS + 1];
 
     // abort ggml_metal_graph_compute if callback returns true
     ggml_abort_callback abort_callback;
@@ -303,7 +329,7 @@ static void * ggml_metal_host_malloc(size_t n) {
     return data;
 }
 
-static struct ggml_backend_metal_context * ggml_metal_init(int n_cb) {
+static struct ggml_backend_metal_context * ggml_metal_init(void) {
     GGML_METAL_LOG_INFO("%s: allocating\n", __func__);
 
 #if TARGET_OS_OSX && !GGML_METAL_NDEBUG
@@ -322,8 +348,9 @@ static struct ggml_backend_metal_context * ggml_metal_init(int n_cb) {
     // Configure context
     struct ggml_backend_metal_context * ctx = calloc(1, sizeof(struct ggml_backend_metal_context));
     ctx->device = device;
-    ctx->n_cb   = MIN(n_cb, GGML_METAL_MAX_BUFFERS);
     ctx->queue  = [ctx->device newCommandQueue];
+    ctx->edesc  = MTLComputePassDescriptor.computePassDescriptor;
+    ctx->edesc.dispatchType = MTLDispatchTypeSerial;
     ctx->d_queue = dispatch_queue_create("ggml-metal", DISPATCH_QUEUE_CONCURRENT);
 
     id<MTLLibrary> metal_library;
@@ -455,7 +482,15 @@ static struct ggml_backend_metal_context * ggml_metal_init(int n_cb) {
     GGML_METAL_LOG_INFO("%s: simdgroup matrix mul. support = %s\n",       __func__, ctx->support_simdgroup_mm ? "true" : "false");
     GGML_METAL_LOG_INFO("%s: hasUnifiedMemory              = %s\n",       __func__, ctx->device.hasUnifiedMemory ? "true" : "false");
 
-    ctx->should_capture_next_compute = false;
+    ctx->capture_next_compute = false;
+    ctx->capture_started = false;
+    ctx->capture_scope = nil;
+
+    ctx->gf = nil;
+    ctx->encode_async = nil;
+    for (int i = 0; i < GGML_METAL_MAX_COMMAND_BUFFERS; ++i) {
+        ctx->command_buffers[i] = nil;
+    }
 
 #if TARGET_OS_OSX || (TARGET_OS_IOS && __clang_major__ >= 15)
     if (@available(macOS 10.12, iOS 16.0, *)) {
@@ -686,6 +721,7 @@ static struct ggml_backend_metal_context * ggml_metal_init(int n_cb) {
     }
 
     [metal_library release];
+
     return ctx;
 }
 
@@ -874,78 +910,23 @@ static bool ggml_metal_supports_op(const struct ggml_backend_metal_context * ctx
     }
 }
 
-static enum ggml_status ggml_metal_graph_compute(
+static void ggml_metal_encode_node(
      struct ggml_backend_metal_context * ctx,
-               struct ggml_cgraph * gf) {
+                                   int   idx,
+          id<MTLComputeCommandEncoder>   encoder) {
+    struct ggml_cgraph * gf = ctx->gf;
 
-    @autoreleasepool {
+    struct ggml_tensor * node = ggml_graph_node(gf, idx);
-    MTLComputePassDescriptor * edesc = MTLComputePassDescriptor.computePassDescriptor;
-    edesc.dispatchType = MTLDispatchTypeSerial;
 
-    // create multiple command buffers and enqueue them
+    //GGML_METAL_LOG_INFO("%s: encoding node %3d, op = %8s\n", __func__, idx, ggml_op_name(node->op));
-    // then, we encode the graph into the command buffers in parallel
 
-    const int n_nodes = gf->n_nodes;
+    struct ggml_tensor * src0 = node->src[0];
-    const int n_cb = ctx->n_cb;
+    struct ggml_tensor * src1 = node->src[1];
-    const int n_nodes_per_cb = (n_nodes + n_cb - 1) / n_cb;
+    struct ggml_tensor * src2 = node->src[2];
-
+    struct ggml_tensor * dst  = node;
-    const bool should_capture = ctx->should_capture_next_compute;
-    if (should_capture) {
-        ctx->should_capture_next_compute = false;
-
-        MTLCaptureDescriptor * descriptor = [MTLCaptureDescriptor new];
-        descriptor.captureObject = ctx->queue;
-
-        NSError * error = nil;
-        if (![[MTLCaptureManager sharedCaptureManager] startCaptureWithDescriptor:descriptor error:&error]) {
-            GGML_METAL_LOG_ERROR("%s: error: unable to start capture '%s'\n", __func__, [[error localizedDescription] UTF8String]);
-            GGML_ABORT("capture failed");
-        }
-    }
-
-    id<MTLCommandBuffer> command_buffer_builder[n_cb];
-    for (int cb_idx = 0; cb_idx < n_cb; ++cb_idx) {
-        id<MTLCommandBuffer> command_buffer  = [ctx->queue commandBufferWithUnretainedReferences];
-        command_buffer_builder[cb_idx] = command_buffer;
-
-        // always enqueue the first two command buffers
-        // enqueue all of the command buffers if we don't need to abort
-        if (cb_idx < 2 || ctx->abort_callback == NULL) {
-            [command_buffer enqueue];
-        }
-    }
-
-    const id<MTLCommandBuffer> *command_buffers = command_buffer_builder;
-
-    dispatch_apply(n_cb, ctx->d_queue, ^(size_t iter) {
-        const int cb_idx = iter;
-
-        size_t offs_src0 = 0;
-        size_t offs_src1 = 0;
-        size_t offs_src2 = 0;
-        size_t offs_dst  = 0;
-
-        id<MTLCommandBuffer> command_buffer  = command_buffers[cb_idx];
-        id<MTLComputeCommandEncoder> encoder = [command_buffer computeCommandEncoderWithDescriptor: edesc];
-
-        const int node_start =                                      (cb_idx + 0) * n_nodes_per_cb;
-        const int node_end   = MIN((cb_idx == n_cb - 1) ? n_nodes : (cb_idx + 1) * n_nodes_per_cb, n_nodes);
-
-        for (int i = node_start; i < node_end; ++i) {
-            if (i == -1) {
-                [encoder memoryBarrierWithScope:MTLBarrierScopeBuffers];
-                continue;
-            }
-
-            //GGML_METAL_LOG_INFO("%s: encoding node %3d, op = %8s\n", __func__, i, ggml_op_name(gf->nodes[i]->op));
-
-            struct ggml_tensor * src0 = gf->nodes[i]->src[0];
-            struct ggml_tensor * src1 = gf->nodes[i]->src[1];
-            struct ggml_tensor * src2 = gf->nodes[i]->src[2];
-            struct ggml_tensor * dst  = gf->nodes[i];
 
     if (ggml_is_empty(dst)) {
-                continue;
+        return;
     }
 
     switch (dst->op) {
@@ -956,7 +937,7 @@ static enum ggml_status ggml_metal_graph_compute(
         case GGML_OP_PERMUTE:
             {
                 // noop -> next node
-                    } continue;
+            } return;
         default:
             {
             } break;
@@ -967,10 +948,6 @@ static enum ggml_status ggml_metal_graph_compute(
         GGML_ABORT("unsupported op");
     }
 
-            if (should_capture) {
-                [encoder pushDebugGroup:[NSString stringWithCString:ggml_op_desc(dst) encoding:NSUTF8StringEncoding]];
-            }
-
     const int64_t  ne00 = src0 ? src0->ne[0] : 0;
     const int64_t  ne01 = src0 ? src0->ne[1] : 0;
     const int64_t  ne02 = src0 ? src0->ne[2] : 0;
@@ -1015,6 +992,11 @@ static enum ggml_status ggml_metal_graph_compute(
     const enum ggml_type src1t = src1 ? src1->type : GGML_TYPE_COUNT;
     const enum ggml_type dstt  = dst  ? dst->type  : GGML_TYPE_COUNT;
 
+    size_t offs_src0 = 0;
+    size_t offs_src1 = 0;
+    size_t offs_src2 = 0;
+    size_t offs_dst  = 0;
+
     id<MTLBuffer> id_src0 = src0 ? ggml_metal_get_buffer(src0, &offs_src0) : nil;
     id<MTLBuffer> id_src1 = src1 ? ggml_metal_get_buffer(src1, &offs_src1) : nil;
     id<MTLBuffer> id_src2 = src2 ? ggml_metal_get_buffer(src2, &offs_src2) : nil;
@@ -1039,7 +1021,7 @@ static enum ggml_status ggml_metal_graph_compute(
             {
                 id<MTLComputePipelineState> pipeline = ctx->kernels[GGML_METAL_KERNEL_TYPE_CONCAT].pipeline;
 
-                        const int32_t dim = ((int32_t *) dst->op_params)[0];
+                const int32_t dim = ((const int32_t *) dst->op_params)[0];
 
                 [encoder setComputePipelineState:pipeline];
                 [encoder setBuffer:id_src0 offset:offs_src0 atIndex:0];
@@ -1203,12 +1185,12 @@ static enum ggml_status ggml_metal_graph_compute(
                 GGML_ASSERT(ggml_is_contiguous(src0));
                 GGML_ASSERT(ggml_is_contiguous(src1));
 
-                        const size_t pnb1 = ((int32_t *) dst->op_params)[0];
+                const size_t pnb1 = ((const int32_t *) dst->op_params)[0];
-                        const size_t pnb2 = ((int32_t *) dst->op_params)[1];
+                const size_t pnb2 = ((const int32_t *) dst->op_params)[1];
-                        const size_t pnb3 = ((int32_t *) dst->op_params)[2];
+                const size_t pnb3 = ((const int32_t *) dst->op_params)[2];
-                        const size_t offs = ((int32_t *) dst->op_params)[3];
+                const size_t offs = ((const int32_t *) dst->op_params)[3];
 
-                        const bool inplace = (bool) ((int32_t *) dst->op_params)[4];
+                const bool inplace = (bool) ((const int32_t *) dst->op_params)[4];
 
                 if (!inplace) {
                     // run a separete kernel to cpy src->dst
@@ -1309,8 +1291,8 @@ static enum ggml_status ggml_metal_graph_compute(
 
                 float min;
                 float max;
-                        memcpy(&min, ((int32_t *) dst->op_params) + 0, sizeof(float));
+                memcpy(&min, ((const int32_t *) dst->op_params) + 0, sizeof(float));
-                        memcpy(&max, ((int32_t *) dst->op_params) + 1, sizeof(float));
+                memcpy(&max, ((const int32_t *) dst->op_params) + 1, sizeof(float));
 
                 [encoder setComputePipelineState:pipeline];
                 [encoder setBuffer:id_src0   offset:offs_src0 atIndex:0];
@@ -1323,7 +1305,7 @@ static enum ggml_status ggml_metal_graph_compute(
                 [encoder dispatchThreadgroups:MTLSizeMake(n, 1, 1) threadsPerThreadgroup:MTLSizeMake(1, 1, 1)];
             } break;
         case GGML_OP_UNARY:
-                    switch (ggml_get_unary_op(gf->nodes[i])) {
+            switch (ggml_get_unary_op(node)) {
                 // we are not taking into account the strides, so for now require contiguous tensors
                 GGML_ASSERT(ggml_is_contiguous(src0));
 
@@ -1422,7 +1404,7 @@ static enum ggml_status ggml_metal_graph_compute(
                 } break;
                 default:
                 {
-                                GGML_METAL_LOG_WARN("%s: node %3d, op = %8s not implemented\n", __func__, i, ggml_op_name(dst->op));
+                    GGML_METAL_LOG_WARN("%s: node %3d, op = %8s not implemented\n", __func__, idx, ggml_op_name(dst->op));
                     GGML_ABORT("fatal error");
                 }
             } break;
@@ -1551,8 +1533,8 @@ static enum ggml_status ggml_metal_graph_compute(
                 float scale;
                 float max_bias;
 
-                        memcpy(&scale,    ((int32_t *) dst->op_params) + 0, sizeof(scale));
+                memcpy(&scale,    ((const int32_t *) dst->op_params) + 0, sizeof(scale));
-                        memcpy(&max_bias, ((int32_t *) dst->op_params) + 1, sizeof(max_bias));
+                memcpy(&max_bias, ((const int32_t *) dst->op_params) + 1, sizeof(max_bias));
 
                 const int64_t nrows_x = ggml_nrows(src0);
                 const int64_t nrows_y = src0->ne[1];
@@ -1585,7 +1567,7 @@ static enum ggml_status ggml_metal_graph_compute(
             } break;
         case GGML_OP_DIAG_MASK_INF:
             {
-                        const int n_past = ((int32_t *)(dst->op_params))[0];
+                const int n_past = ((const int32_t *)(dst->op_params))[0];
 
                 id<MTLComputePipelineState> pipeline = nil;
 
@@ -1644,9 +1626,9 @@ static enum ggml_status ggml_metal_graph_compute(
             } break;
         case GGML_OP_SSM_SCAN:
             {
-                        struct ggml_tensor * src3 = gf->nodes[i]->src[3];
+                struct ggml_tensor * src3 = node->src[3];
-                        struct ggml_tensor * src4 = gf->nodes[i]->src[4];
+                struct ggml_tensor * src4 = node->src[4];
-                        struct ggml_tensor * src5 = gf->nodes[i]->src[5];
+                struct ggml_tensor * src5 = node->src[5];
 
                 GGML_ASSERT(src3);
                 GGML_ASSERT(src4);
@@ -2425,7 +2407,7 @@ static enum ggml_status ggml_metal_graph_compute(
                 float eps;
                 memcpy(&eps, dst->op_params + 1, sizeof(float));
 
-                        const int32_t n_groups = ((int32_t *) dst->op_params)[0];
+                const int32_t n_groups = ((const int32_t *) dst->op_params)[0];
 
                 int nth = 32; // SIMD width
 
@@ -2479,11 +2461,11 @@ static enum ggml_status ggml_metal_graph_compute(
 
                 const int nth = MIN(1024, ne00);
 
-                        const int n_past     = ((int32_t *) dst->op_params)[0];
+                const int n_past     = ((const int32_t *) dst->op_params)[0];
-                        const int n_dims     = ((int32_t *) dst->op_params)[1];
+                const int n_dims     = ((const int32_t *) dst->op_params)[1];
-                        const int mode       = ((int32_t *) dst->op_params)[2];
+                const int mode       = ((const int32_t *) dst->op_params)[2];
                 // skip 3, n_ctx, used in GLM RoPE, unimplemented in metal
-                        const int n_ctx_orig = ((int32_t *) dst->op_params)[4];
+                const int n_ctx_orig = ((const int32_t *) dst->op_params)[4];
 
                 float freq_base;
                 float freq_scale;
@@ -2492,12 +2474,12 @@ static enum ggml_status ggml_metal_graph_compute(
                 float beta_fast;
                 float beta_slow;
 
-                        memcpy(&freq_base,   (int32_t *) dst->op_params +  5, sizeof(float));
+                memcpy(&freq_base,   (const int32_t *) dst->op_params +  5, sizeof(float));
-                        memcpy(&freq_scale,  (int32_t *) dst->op_params +  6, sizeof(float));
+                memcpy(&freq_scale,  (const int32_t *) dst->op_params +  6, sizeof(float));
-                        memcpy(&ext_factor,  (int32_t *) dst->op_params +  7, sizeof(float));
+                memcpy(&ext_factor,  (const int32_t *) dst->op_params +  7, sizeof(float));
-                        memcpy(&attn_factor, (int32_t *) dst->op_params +  8, sizeof(float));
+                memcpy(&attn_factor, (const int32_t *) dst->op_params +  8, sizeof(float));
-                        memcpy(&beta_fast,   (int32_t *) dst->op_params +  9, sizeof(float));
+                memcpy(&beta_fast,   (const int32_t *) dst->op_params +  9, sizeof(float));
-                        memcpy(&beta_slow,   (int32_t *) dst->op_params + 10, sizeof(float));
+                memcpy(&beta_slow,   (const int32_t *) dst->op_params + 10, sizeof(float));
 
                 const bool is_neox = mode & GGML_ROPE_TYPE_NEOX;
 
@@ -2686,8 +2668,8 @@ static enum ggml_status ggml_metal_graph_compute(
                 float start;
                 float step;
 
-                        memcpy(&start, ((int32_t *) dst->op_params) + 0, sizeof(float));
+                memcpy(&start, ((const int32_t *) dst->op_params) + 0, sizeof(float));
-                        memcpy(&step,  ((int32_t *) dst->op_params) + 2, sizeof(float));
+                memcpy(&step,  ((const int32_t *) dst->op_params) + 2, sizeof(float));
 
                 id<MTLComputePipelineState> pipeline = ctx->kernels[GGML_METAL_KERNEL_TYPE_ARANGE_F32].pipeline;
 
@@ -2786,7 +2768,7 @@ static enum ggml_status ggml_metal_graph_compute(
 
                 GGML_ASSERT(ggml_are_same_shape (src1, src2));
 
-                        struct ggml_tensor * src3 = gf->nodes[i]->src[3];
+                struct ggml_tensor * src3 = node->src[3];
 
                 size_t offs_src3 = 0;
 
@@ -2811,9 +2793,9 @@ static enum ggml_status ggml_metal_graph_compute(
                 float scale;
                 float max_bias;
                 float logit_softcap;
-                        memcpy(&scale,         ((int32_t *) dst->op_params) + 0, sizeof(scale));
+                memcpy(&scale,         ((const int32_t *) dst->op_params) + 0, sizeof(scale));
-                        memcpy(&max_bias,      ((int32_t *) dst->op_params) + 1, sizeof(max_bias));
+                memcpy(&max_bias,      ((const int32_t *) dst->op_params) + 1, sizeof(max_bias));
-                        memcpy(&logit_softcap, ((int32_t *) dst->op_params) + 2, sizeof(logit_softcap));
+                memcpy(&logit_softcap, ((const int32_t *) dst->op_params) + 2, sizeof(logit_softcap));
 
                 if (logit_softcap != 0.0f) {
                     scale /= logit_softcap;
@@ -3014,10 +2996,87 @@ static enum ggml_status ggml_metal_graph_compute(
             } break;
        default:
             {
-                        GGML_METAL_LOG_ERROR("%s: error: node %3d, op = %8s not implemented\n", __func__, i, ggml_op_name(dst->op));
+                GGML_METAL_LOG_ERROR("%s: error: node %3d, op = %8s not implemented\n", __func__, idx, ggml_op_name(dst->op));
                 GGML_ABORT("fatal error");
             }
     }
+}
+
+static enum ggml_status ggml_metal_graph_compute(
+        struct ggml_backend_metal_context * ctx,
+                       struct ggml_cgraph * gf) {
+    // number of nodes encoded by the main thread (empirically determined)
+    const int n_main = 128;
+
+    // number of threads in addition to the main thread
+    const int n_cb = ctx->n_cb;
+
+    // submit the ggml compute graph to the GPU by creating command buffers and encoding the ops in them
+    // the first n_nodes_0 are encoded and submitted for processing directly by the calling thread
+    // while these nodes are processing, we start n_cb threads to enqueue the rest of the nodes
+    // each thread creates it's own command buffer and enqueues the ops in parallel
+    //
+    // tests on M1 Pro and M2 Ultra using LLaMA models, show that optimal values for n_cb are 1 or 2
+
+    @autoreleasepool {
+        ctx->gf = gf;
+
+        ctx->n_nodes_0 = MIN(n_main, gf->n_nodes);
+        ctx->n_nodes_1 = gf->n_nodes - ctx->n_nodes_0;
+
+        ctx->n_nodes_per_cb = (ctx->n_nodes_1 + ctx->n_cb - 1) / ctx->n_cb;
+
+        const bool should_capture = ctx->capture_next_compute;
+        if (should_capture) {
+            ctx->capture_next_compute = false;
+
+            if (!ctx->capture_started) {
+                // create capture scope
+                ctx->capture_scope = [[MTLCaptureManager sharedCaptureManager] newCaptureScopeWithDevice:ctx->device];
+
+                MTLCaptureDescriptor * descriptor = [MTLCaptureDescriptor new];
+                descriptor.captureObject = ctx->capture_scope;
+                descriptor.destination = MTLCaptureDestinationGPUTraceDocument;
+                descriptor.outputURL = [NSURL fileURLWithPath:[NSString stringWithFormat:@"/tmp/perf-metal.gputrace"]];
+
+                NSError * error = nil;
+                if (![[MTLCaptureManager sharedCaptureManager] startCaptureWithDescriptor:descriptor error:&error]) {
+                    GGML_METAL_LOG_ERROR("%s: error: unable to start capture '%s'\n", __func__, [[error localizedDescription] UTF8String]);
+                    GGML_ABORT("capture failed");
+                } else {
+                    [ctx->capture_scope beginScope];
+                    ctx->capture_started = true;
+                }
+            }
+        }
+
+        // TODO: how to avoid this allocation? I tried initializing it in ggml_backend_metal_set_n_cb but it crashes.
+        ctx->encode_async = ^(size_t iter) {
+            const int cb_idx = iter;
+            const int n_cb_l = ctx->n_cb;
+
+            const int n_nodes_0 = ctx->n_nodes_0;
+            const int n_nodes_1 = ctx->n_nodes_1;
+
+            const int n_nodes_per_cb = ctx->n_nodes_per_cb;
+
+            id<MTLCommandBuffer> command_buffer  = ctx->command_buffers[cb_idx];
+            id<MTLComputeCommandEncoder> encoder = [command_buffer computeCommandEncoderWithDescriptor: ctx->edesc];
+
+            int node_start = 0;
+            int node_end   = n_nodes_0;
+
+            if (cb_idx < n_cb_l) {
+                node_start = n_nodes_0 + (                                         (cb_idx + 0) * n_nodes_per_cb);
+                node_end   = n_nodes_0 + (MIN((cb_idx == n_cb_l - 1) ? n_nodes_1 : (cb_idx + 1) * n_nodes_per_cb, n_nodes_1));
+            }
+
+            for (int idx = node_start; idx < node_end; ++idx) {
+                if (should_capture) {
+                    [encoder pushDebugGroup:[NSString stringWithCString:ggml_op_desc(ggml_graph_node(gf, idx)) encoding:NSUTF8StringEncoding]];
+                }
+
+                ggml_metal_encode_node(ctx, idx, encoder);
 
                 if (should_capture) {
                     [encoder popDebugGroup];
@@ -3029,13 +3088,52 @@ static enum ggml_status ggml_metal_graph_compute(
             if (cb_idx < 2 || ctx->abort_callback == NULL) {
                 [command_buffer commit];
             }
-    });
+        };
 
-    // Wait for completion and check status of each command buffer
+        // the main thread commits the first few commands immediately
+        // command_buffer[n_cb]
+        {
+            id<MTLCommandBuffer> command_buffer = [ctx->queue commandBufferWithUnretainedReferences];
+            ctx->command_buffers[n_cb] = command_buffer;
+
+            [command_buffer enqueue];
+            ctx->encode_async(n_cb);
+        }
+
+        // prepare the rest of the command buffers asynchronously
+        // command_buffer[0.. n_cb)
+        for (int cb_idx = 0; cb_idx < n_cb; ++cb_idx) {
+            id<MTLCommandBuffer> command_buffer = [ctx->queue commandBufferWithUnretainedReferences];
+            ctx->command_buffers[cb_idx] = command_buffer;
+
+            // always enqueue the first two command buffers
+            // enqueue all of the command buffers if we don't need to abort
+            if (cb_idx < 2 || ctx->abort_callback == NULL) {
+                [command_buffer enqueue];
+            }
+        }
+
+        dispatch_apply(n_cb, ctx->d_queue, ctx->encode_async);
+
+        // wait for completion and check status of each command buffer
         // needed to detect if the device ran out-of-memory for example (#1881)
+        {
+            id<MTLCommandBuffer> command_buffer = ctx->command_buffers[n_cb];
+            [command_buffer waitUntilCompleted];
+
+            MTLCommandBufferStatus status = [command_buffer status];
+            if (status != MTLCommandBufferStatusCompleted) {
+                GGML_METAL_LOG_INFO("%s: command buffer %d failed with status %lu\n", __func__, n_cb, status);
+                if (status == MTLCommandBufferStatusError) {
+                    GGML_METAL_LOG_INFO("error: %s\n", [[command_buffer error].localizedDescription UTF8String]);
+                }
+
+                return GGML_STATUS_FAILED;
+            }
+        }
 
         for (int i = 0; i < n_cb; ++i) {
-        id<MTLCommandBuffer> command_buffer = command_buffers[i];
+            id<MTLCommandBuffer> command_buffer = ctx->command_buffers[i];
             [command_buffer waitUntilCompleted];
 
             MTLCommandBufferStatus status = [command_buffer status];
@@ -3048,12 +3146,12 @@ static enum ggml_status ggml_metal_graph_compute(
                 return GGML_STATUS_FAILED;
             }
 
-        id<MTLCommandBuffer> next_buffer = (i + 1 < n_cb ? command_buffers[i + 1] : nil);
+            id<MTLCommandBuffer> next_buffer = (i + 1 < n_cb ? ctx->command_buffers[i + 1] : nil);
             if (!next_buffer) {
                 continue;
             }
 
-        bool next_queued = ([next_buffer status] != MTLCommandBufferStatusNotEnqueued);
+            const bool next_queued = ([next_buffer status] != MTLCommandBufferStatusNotEnqueued);
             if (next_queued) {
                 continue;
             }
@@ -3066,11 +3164,12 @@ static enum ggml_status ggml_metal_graph_compute(
             [next_buffer commit];
         }
 
-    if (should_capture) {
+        if (!should_capture && ctx->capture_started) {
+            [ctx->capture_scope endScope];
             [[MTLCaptureManager sharedCaptureManager] stopCapture];
         }
-
     }
+
     return GGML_STATUS_SUCCESS;
 }
 
@@ -3167,6 +3266,7 @@ static struct ggml_backend_buffer_i ggml_backend_metal_buffer_i = {
     /* .free_buffer     = */ ggml_backend_metal_buffer_free_buffer,
     /* .get_base        = */ ggml_backend_metal_buffer_get_base,
     /* .init_tensor     = */ NULL,
+    /* .memset_tensor   = */ NULL,
     /* .set_tensor      = */ ggml_backend_metal_buffer_set_tensor,
     /* .get_tensor      = */ ggml_backend_metal_buffer_get_tensor,
     /* .cpy_tensor      = */ ggml_backend_metal_buffer_cpy_tensor,
@@ -3404,6 +3504,25 @@ GGML_CALL static bool ggml_backend_metal_supports_buft(ggml_backend_t backend, g
     UNUSED(backend);
 }
 
+static void ggml_backend_metal_set_n_cb(ggml_backend_t backend, int n_cb) {
+    GGML_ASSERT(ggml_backend_is_metal(backend));
+
+    struct ggml_backend_metal_context * ctx = (struct ggml_backend_metal_context *)backend->context;
+
+    if (ctx->n_cb != n_cb) {
+        ctx->n_cb = MIN(n_cb, GGML_METAL_MAX_COMMAND_BUFFERS);
+
+        if (ctx->n_cb > 2) {
+            GGML_METAL_LOG_WARN("%s: n_cb = %d, using n_cb > 2 is not recommended and can degrade the performance in some cases\n", __func__, n_cb);
+        }
+    }
+
+    // TODO: setting encode_async here causes crash during the next ggml_metal_graph_compute call. why?
+    //ctx->encode_async = ^(size_t iter) {
+    //    ...
+    //};
+}
+
 static struct ggml_backend_i ggml_backend_metal_i = {
     /* .get_name                = */ ggml_backend_metal_name,
     /* .free                    = */ ggml_backend_metal_free,
@@ -3438,35 +3557,29 @@ static ggml_guid_t ggml_backend_metal_guid(void) {
 }
 
 ggml_backend_t ggml_backend_metal_init(void) {
-    struct ggml_backend_metal_context * ctx = ggml_metal_init(GGML_DEFAULT_N_THREADS);
+    struct ggml_backend_metal_context * ctx = ggml_metal_init();
     if (ctx == NULL) {
         GGML_METAL_LOG_ERROR("%s: error: failed to allocate context\n", __func__);
         return NULL;
     }
 
-    ggml_backend_t metal_backend = malloc(sizeof(struct ggml_backend));
+    ggml_backend_t backend = malloc(sizeof(struct ggml_backend));
 
-    *metal_backend = (struct ggml_backend) {
+    *backend = (struct ggml_backend) {
         /* .guid      = */ ggml_backend_metal_guid(),
         /* .interface = */ ggml_backend_metal_i,
         /* .context   = */ ctx,
     };
 
-    return metal_backend;
+    ggml_backend_metal_set_n_cb(backend, 1);
+
+    return backend;
 }
 
 bool ggml_backend_is_metal(ggml_backend_t backend) {
     return backend != NULL && ggml_guid_matches(backend->guid, ggml_backend_metal_guid());
 }
 
-void ggml_backend_metal_set_n_cb(ggml_backend_t backend, int n_cb) {
-    GGML_ASSERT(ggml_backend_is_metal(backend));
-
-    struct ggml_backend_metal_context * ctx = (struct ggml_backend_metal_context *)backend->context;
-
-    ctx->n_cb = MIN(n_cb, GGML_METAL_MAX_BUFFERS);
-}
-
 void ggml_backend_metal_set_abort_callback(ggml_backend_t backend, ggml_abort_callback abort_callback, void * user_data) {
     GGML_ASSERT(ggml_backend_is_metal(backend));
 
@@ -3488,7 +3601,7 @@ void ggml_backend_metal_capture_next_compute(ggml_backend_t backend) {
     GGML_ASSERT(ggml_backend_is_metal(backend));
 
     struct ggml_backend_metal_context * ctx = (struct ggml_backend_metal_context *)backend->context;
-    ctx->should_capture_next_compute = true;
+    ctx->capture_next_compute = true;
 }
 
 GGML_CALL ggml_backend_t ggml_backend_reg_metal_init(const char * params, void * user_data); // silence warning

ggml/src/ggml-metal.metal

@@ -2631,11 +2631,11 @@ kernel void kernel_flash_attn_ext_vec_f16(
         const short iv3 = iq3 / rv3;
 
         // load the queries from shared memory into local memory
-        half4 mq[D4];
+        float4 mq[D4];
 
         for (short ii = 0; ii < D4; ii += NW) {
             short i = ii + tiisg;
-            mq[i] = sq4[i];
+            mq[i] = (float4) sq4[i];
         }
 
         // pointer to the mask
@@ -2661,11 +2661,11 @@ kernel void kernel_flash_attn_ext_vec_f16(
                     for (short ii = 0; ii < D4; ii += NW) {
                         const short i = ii + tiisg;
 
-                        half4x4 mk;
+                        float4x4 mk;
-                        mk[0] = pk4[i + 0*(nb11/8)];
+                        mk[0] = (float4) pk4[i + 0*(nb11/8)];
-                        mk[1] = pk4[i + 1*(nb11/8)];
+                        mk[1] = (float4) pk4[i + 1*(nb11/8)];
-                        mk[2] = pk4[i + 2*(nb11/8)];
+                        mk[2] = (float4) pk4[i + 2*(nb11/8)];
-                        mk[3] = pk4[i + 3*(nb11/8)];
+                        mk[3] = (float4) pk4[i + 3*(nb11/8)];
 
                         mqk += (float4) (mq[i] * mk);
                     }

ggml/src/ggml-quants.c

@@ -4013,7 +4013,7 @@ void ggml_vec_dot_q4_0_q8_0(int n, float * restrict s, size_t bs, const void * r
     svfloat32_t sumv0 = svdup_n_f32(0.0f);
     svfloat32_t sumv1 = svdup_n_f32(0.0f);
 
-    const int vector_length = ggml_sve_cnt_b*8;
+    const int vector_length = ggml_cpu_get_sve_cnt()*8;
 
     // VLA Implementation using switch case
     switch (vector_length) {
@@ -5597,7 +5597,7 @@ void ggml_vec_dot_q8_0_q8_0(int n, float * restrict s, size_t bs, const void * r
     svfloat32_t sumv0 = svdup_n_f32(0.0f);
     svfloat32_t sumv1 = svdup_n_f32(0.0f);
 
-    const int vector_length = ggml_sve_cnt_b*8;
+    const int vector_length = ggml_cpu_get_sve_cnt()*8;
 
     //VLA Implemenation for SVE
     switch (vector_length) {

ggml/src/ggml-quants.h

@@ -142,10 +142,6 @@ void iq2xs_free_impl(enum ggml_type type);
 void iq3xs_init_impl(int grid_size);
 void iq3xs_free_impl(int grid_size);
 
-#if defined(__ARM_FEATURE_SVE)
-extern int ggml_sve_cnt_b;
-#endif
-
 #ifdef __cplusplus
 }
 #endif

ggml/src/ggml-rpc.cpp

@@ -469,6 +469,7 @@ static ggml_backend_buffer_i ggml_backend_rpc_buffer_interface = {
     /* .free_buffer     = */ ggml_backend_rpc_buffer_free_buffer,
     /* .get_base        = */ ggml_backend_rpc_buffer_get_base,
     /* .init_tensor     = */ ggml_backend_rpc_buffer_init_tensor,
+    /* .memset_tensor   = */ NULL,
     /* .set_tensor      = */ ggml_backend_rpc_buffer_set_tensor,
     /* .get_tensor      = */ ggml_backend_rpc_buffer_get_tensor,
     /* .cpy_tensor      = */ ggml_backend_rpc_buffer_cpy_tensor,

ggml/src/ggml-sycl.cpp

@@ -3496,8 +3496,7 @@ static void ggml_sycl_mul_mat(ggml_backend_sycl_context & ctx, const ggml_tensor
 
     bool use_mul_mat_vec_q =  ggml_is_quantized(src0->type)
         && src1->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32
-        && src1->ne[1] <= MMVQ_MAX_BATCH_SIZE
+        && src1->ne[1] <= MMVQ_MAX_BATCH_SIZE;
-        && (ctx.stream()->get_backend() == sycl::backend::ext_oneapi_cuda || src1->ne[1] > MMVQ_MIN_BATCH_SIZE);
 
     bool use_mul_mat_q =  ggml_sycl_supports_mmq(src0->type)
         && src1->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32;
@@ -4323,6 +4322,7 @@ static struct ggml_backend_buffer_i ggml_backend_sycl_buffer_interface = {
     /* .free_buffer     = */ ggml_backend_sycl_buffer_free_buffer,
     /* .get_base        = */ ggml_backend_sycl_buffer_get_base,
     /* .init_tensor     = */ ggml_backend_sycl_buffer_init_tensor,
+    /* .memset_tensor   = */ NULL,
     /* .set_tensor      = */ ggml_backend_sycl_buffer_set_tensor,
     /* .get_tensor      = */ ggml_backend_sycl_buffer_get_tensor,
     /* .cpy_tensor      = */ ggml_backend_sycl_buffer_cpy_tensor,
@@ -4734,6 +4734,7 @@ static struct ggml_backend_buffer_i ggml_backend_sycl_split_buffer_interface = {
     /* .free_buffer     = */ ggml_backend_sycl_split_buffer_free_buffer,
     /* .get_base        = */ ggml_backend_sycl_split_buffer_get_base,
     /* .init_tensor     = */ ggml_backend_sycl_split_buffer_init_tensor,
+    /* .memset_tensor   = */ NULL,
     /* .set_tensor      = */ ggml_backend_sycl_split_buffer_set_tensor,
     /* .get_tensor      = */ ggml_backend_sycl_split_buffer_get_tensor,
     /* .cpy_tensor      = */ NULL,

ggml/src/ggml-sycl/common.hpp

@@ -134,7 +134,6 @@ typedef sycl::float2 dfloat2;
 #endif // GGML_SYCL_F16
 
 #define MMVQ_MAX_BATCH_SIZE  8
-#define MMVQ_MIN_BATCH_SIZE  4
 
 static const int8_t kvalues_iq4nl[16]={-127, -104, -83, -65, -49, -35, -22, -10, 1, 13, 25, 38, 53, 69, 89, 113};
 

ggml/src/ggml-vulkan.cpp

@@ -20,6 +20,8 @@
 #include <unordered_map>
 #include <memory>
 #include <mutex>
+#include <future>
+#include <thread>
 
 #include "ggml-impl.h"
 #include "ggml-backend-impl.h"
@@ -607,13 +609,16 @@ typedef void (*ggml_vk_func_t)(ggml_backend_vk_context * ctx, vk_context& subctx
 
 GGML_CALL static void ggml_backend_vk_free(ggml_backend_t backend);
 
-static void ggml_vk_create_pipeline(vk_device& device, vk_pipeline& pipeline, const std::string& name, size_t spv_size, const void* spv_data, const std::string& entrypoint, uint32_t parameter_count, uint32_t push_constant_size, std::array<uint32_t, 3> wg_denoms, std::vector<uint32_t>&& specialization_constants, uint32_t align) {
+// variables to track number of compiles in progress
+static uint32_t compile_count = 0;
+static std::mutex compile_count_mutex;
+static std::condition_variable compile_count_cond;
+
+static void ggml_vk_create_pipeline_func(vk_device& device, vk_pipeline& pipeline, const std::string name, size_t spv_size, const void* spv_data, const std::string entrypoint, uint32_t parameter_count, uint32_t push_constant_size, std::array<uint32_t, 3> wg_denoms, std::vector<uint32_t> specialization_constants, uint32_t align) {
     VK_LOG_DEBUG("ggml_vk_create_pipeline(" << device->name << ", " << name << ", " << entrypoint << ", " << parameter_count << ", " << push_constant_size << ", (" << wg_denoms[0] << "," << wg_denoms[1] << "," << wg_denoms[2] << "), specialization_constants, " << align << ")");
     GGML_ASSERT(parameter_count > 0);
     GGML_ASSERT(wg_denoms[0] > 0 && wg_denoms[1] > 0 && wg_denoms[2] > 0); // NOLINT
 
-    std::lock_guard<std::mutex> guard(device->mutex);
-
     pipeline = std::make_shared<vk_pipeline_struct>();
     pipeline->name = name;
     pipeline->parameter_count = parameter_count;
@@ -681,7 +686,17 @@ static void ggml_vk_create_pipeline(vk_device& device, vk_pipeline& pipeline, co
         pipeline->layout);
     pipeline->pipeline = device->device.createComputePipeline(VK_NULL_HANDLE, compute_pipeline_create_info).value;
 
+    {
+        std::lock_guard<std::mutex> guard(device->mutex);
         device->pipelines.insert({ pipeline->name, pipeline });
+    }
+
+    {
+        std::lock_guard<std::mutex> guard(compile_count_mutex);
+        assert(compile_count > 0);
+        compile_count--;
+    }
+    compile_count_cond.notify_all();
 }
 
 static void ggml_vk_destroy_pipeline(vk::Device& device, vk_pipeline& pipeline) {
@@ -1079,7 +1094,8 @@ static vk_buffer ggml_vk_create_buffer_device(vk_device& device, size_t size) {
             // Fall back to host memory type
             buf = ggml_vk_create_buffer(device, size, vk::MemoryPropertyFlagBits::eDeviceLocal, vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent);
         } else {
-            buf = ggml_vk_create_buffer(device, size, vk::MemoryPropertyFlagBits::eDeviceLocal);
+            // use rebar if available, otherwise fallback to device only visible memory
+            buf = ggml_vk_create_buffer(device, size, vk::MemoryPropertyFlagBits::eDeviceLocal | vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent, vk::MemoryPropertyFlagBits::eDeviceLocal);
         }
     } catch (const vk::SystemError& e) {
         std::cerr << "ggml_vulkan: Device memory allocation of size " << size << " failed." << std::endl;
@@ -1148,11 +1164,11 @@ static void ggml_vk_load_shaders(vk_device& device) {
     // mulmat
     std::initializer_list<uint32_t> warptile_l = { 128, 128, 128, 16, device->subgroup_size * 2, 64, 2, 4, 4, device->subgroup_size };
     std::initializer_list<uint32_t> warptile_m = { 128,  64,  64, 16, device->subgroup_size, 32, 2, 4, 2, device->subgroup_size };
-    std::initializer_list<uint32_t> warptile_s = { device->subgroup_size,  32,  32, 16, 32, 32, 2, 2, 2, device->subgroup_size };
+    std::initializer_list<uint32_t> warptile_s = { std::max(device->subgroup_size, 16u),  32,  32, 16, 32, 32, 2, 2, 2, device->subgroup_size };
 
     std::initializer_list<uint32_t> warptile_mmq_l = { 128, 128, 128, 32, device->subgroup_size * 2, 64, 2, 4, 4, device->subgroup_size };
     std::initializer_list<uint32_t> warptile_mmq_m = { 128,  64,  64, 32, device->subgroup_size, 32, 2, 4, 2, device->subgroup_size };
-    std::initializer_list<uint32_t> warptile_mmq_s = { device->subgroup_size,  32,  32, 32, 32, 32, 2, 2, 2, device->subgroup_size };
+    std::initializer_list<uint32_t> warptile_mmq_s = { std::max(device->subgroup_size, 16u),  32,  32, 32, 32, 32, 2, 2, 2, device->subgroup_size };
 
     std::array<uint32_t, 3> l_wg_denoms = {128, 128, 1 };
     std::array<uint32_t, 3> m_wg_denoms = { 64,  64, 1 };
@@ -1193,6 +1209,20 @@ static void ggml_vk_load_shaders(vk_device& device) {
     device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K] = std::make_shared<vk_matmul_pipeline_struct>();
     device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_IQ4_NL] = std::make_shared<vk_matmul_pipeline_struct>();
 
+    std::vector<std::future<void>> compiles;
+    auto const &ggml_vk_create_pipeline = [&](vk_device& device, vk_pipeline& pipeline, const std::string &name, size_t spv_size, const void* spv_data, const std::string &entrypoint, uint32_t parameter_count, uint32_t push_constant_size, std::array<uint32_t, 3> wg_denoms, std::vector<uint32_t>&& specialization_constants, uint32_t align) {
+        {
+            // wait until fewer than N compiles are in progress
+            uint32_t N = std::max(1u, std::thread::hardware_concurrency());
+            std::unique_lock<std::mutex> guard(compile_count_mutex);
+            while (compile_count >= N) {
+                compile_count_cond.wait(guard);
+            }
+            compile_count++;
+        }
+        compiles.push_back(std::async(ggml_vk_create_pipeline_func, std::ref(device), std::ref(pipeline), name, spv_size, spv_data, entrypoint, parameter_count, push_constant_size, wg_denoms, specialization_constants, align));
+    };
+
     if (device->fp16) {
         ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->l, "matmul_f32_l", matmul_f32_f32_len, matmul_f32_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, 1);
         ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->m, "matmul_f32_m", matmul_f32_f32_len, matmul_f32_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, 1);
@@ -1742,6 +1772,10 @@ static void ggml_vk_load_shaders(vk_device& device) {
     ggml_vk_create_pipeline(device, device->pipeline_im2col_f32_f16, "im2col_f32_f16", im2col_f32_f16_len, im2col_f32_f16_data, "main", 2, sizeof(vk_op_im2col_push_constants), {256, 1, 1}, {}, 1);
 
     ggml_vk_create_pipeline(device, device->pipeline_timestep_embedding_f32, "timestep_embedding_f32", timestep_embedding_f32_len, timestep_embedding_f32_data, "main", 2, sizeof(vk_op_timestep_embedding_push_constants), {256, 1, 1}, {}, 1);
+
+    for (auto &c : compiles) {
+        c.wait();
+    }
 }
 
 static vk_device ggml_vk_get_device(size_t idx) {
@@ -2806,7 +2840,11 @@ static void ggml_vk_buffer_read_async(vk_context subctx, vk_buffer& src, size_t
 
 static void ggml_vk_buffer_read(vk_buffer& src, size_t offset, void * dst, size_t size) {
     VK_LOG_DEBUG("ggml_vk_buffer_read(" << src->buffer << ", " << offset << ", " << size << ")");
-    if(src->memory_property_flags & vk::MemoryPropertyFlagBits::eHostVisible) {
+
+    // If the device is not an UMA device the memory is host-accessible through rebar. While writing
+    // through PCIe is sufficient fast reading back data from PCIe is slower than going through
+    // the HW device to host copy path.
+    if(src->memory_property_flags & vk::MemoryPropertyFlagBits::eHostVisible && src->device->uma) {
         GGML_ASSERT(src->memory_property_flags & vk::MemoryPropertyFlagBits::eHostCoherent);
 
         memcpy(dst, (uint8_t *) src->ptr + offset, size);
@@ -5008,6 +5046,8 @@ static void ggml_vk_test_matmul(ggml_backend_vk_context * ctx, size_t m, size_t
         }
     }
 
+    ggml_pipeline_allocate_descriptor_sets(ctx->device);
+
     vk_buffer d_X = ggml_vk_create_buffer_check(ctx->device, sizeof(X_TYPE) * x_ne, vk::MemoryPropertyFlagBits::eDeviceLocal);
     vk_buffer d_Y = ggml_vk_create_buffer_check(ctx->device, sizeof(Y_TYPE) * y_ne, vk::MemoryPropertyFlagBits::eDeviceLocal);
     vk_buffer d_D = ggml_vk_create_buffer_check(ctx->device, sizeof(float) * d_ne, vk::MemoryPropertyFlagBits::eDeviceLocal);
@@ -5124,7 +5164,9 @@ static void ggml_vk_test_matmul(ggml_backend_vk_context * ctx, size_t m, size_t
 
     avg_err /= m * n;
 
-    std::cerr << "TEST " << shname << " m=" << m << " n=" << n << " k=" << k << " batch=" << batch << " split_k=" << split_k << " matmul " << time / num_it << "ms avg_err=" << avg_err << std::endl;
+    double tflops = 2.0*m*n*k*batch*num_it / (time / 1000.0) / (1000.0*1000.0*1000.0*1000.0);
+
+    std::cerr << "TEST " << shname << " m=" << m << " n=" << n << " k=" << k << " batch=" << batch << " split_k=" << split_k << " matmul " << time / num_it << "ms " << tflops << " TFLOPS avg_err=" << avg_err << std::endl;
 
     if (avg_err > 0.1) {
         std::cerr << "m = " << first_err_m << " n = " << first_err_n << " b = " << first_err_b << std::endl;
@@ -5246,12 +5288,14 @@ static void ggml_vk_test_dequant(ggml_backend_vk_context * ctx, size_t ne, ggml_
 
     ggml_pipeline_request_descriptor_sets(ctx->device, p, 1);
 
+    ggml_pipeline_allocate_descriptor_sets(ctx->device);
+
     ggml_vk_buffer_write(qx_buf, 0, qx, qx_sz);
 
     vk_context subctx = ggml_vk_create_context(ctx, ctx->device->compute_queue);
     ggml_vk_ctx_begin(ctx->device, subctx);
     const std::vector<uint32_t> pc = { 1, (uint32_t)ne, (uint32_t)ne, (uint32_t)ne, (uint32_t)ne };
-    ggml_vk_dispatch_pipeline(ctx, subctx, p, { { qx_buf, 0, qx_sz }, { x_buf, 0, x_sz_f16 } }, pc.size() * sizeof(int), pc.data(), { (uint32_t)ne, 1, 1});
+    ggml_vk_dispatch_pipeline(ctx, subctx, p, { vk_subbuffer{ qx_buf, 0, qx_sz }, vk_subbuffer{ x_buf, 0, x_sz_f16 } }, pc.size() * sizeof(int), pc.data(), { (uint32_t)ne, 1, 1});
     ggml_vk_ctx_end(subctx);
 
     auto begin = std::chrono::high_resolution_clock::now();
@@ -5378,6 +5422,8 @@ static void ggml_vk_test_dequant_matmul(ggml_backend_vk_context * ctx, size_t m,
         }
     }
 
+    ggml_pipeline_allocate_descriptor_sets(ctx->device);
+
     ggml_vk_buffer_write(qx_buf, 0, qx, qx_sz);
     ggml_vk_buffer_write(y_buf, 0, y, y_sz);
 
@@ -5445,7 +5491,9 @@ static void ggml_vk_test_dequant_matmul(ggml_backend_vk_context * ctx, size_t m,
 
     avg_err /= m * n;
 
-    std::cerr << "TEST MMQ " << shname << " m=" << m << " n=" << n << " k=" << k << " batch=" << batch << " split_k=" << split_k << " matmul " << time_ms / num_it << "ms avg_err=" << avg_err << std::endl;
+    double tflops = 2.0*m*n*k*batch*num_it / (time_ms / 1000.0) / (1000.0*1000.0*1000.0*1000.0);
+
+    std::cerr << "TEST MMQ " << shname << " m=" << m << " n=" << n << " k=" << k << " batch=" << batch << " split_k=" << split_k << " matmul " << time_ms / num_it << "ms " << tflops << " TFLOPS avg_err=" << avg_err << std::endl;
 
     if (avg_err > 0.01 || std::isnan(avg_err)) {
         std::cerr << "m = " << first_err_m << " n = " << first_err_n << " b = " << first_err_b << std::endl;
@@ -5497,9 +5545,6 @@ static ggml_tensor_extra_gpu * ggml_vk_tensor_create_extra(ggml_tensor * tensor)
 
 static void ggml_vk_preallocate_buffers(ggml_backend_vk_context * ctx) {
 #if defined(GGML_VULKAN_RUN_TESTS)
-    ctx->staging = ggml_vk_create_buffer_check(ctx->device, 100ul * 1024ul * 1024ul,
-        vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent | vk::MemoryPropertyFlagBits::eHostCached,
-        vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent);
     ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_F32);
     ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_Q4_0);
     ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_Q4_1);
@@ -6246,6 +6291,7 @@ static ggml_backend_buffer_i ggml_backend_vk_buffer_interface = {
     /* .free_buffer     = */ ggml_backend_vk_buffer_free_buffer,
     /* .get_base        = */ ggml_backend_vk_buffer_get_base,
     /* .init_tensor     = */ ggml_backend_vk_buffer_init_tensor,
+    /* .memset_tensor   = */ NULL,
     /* .set_tensor      = */ ggml_backend_vk_buffer_set_tensor,
     /* .get_tensor      = */ ggml_backend_vk_buffer_get_tensor,
     /* .cpy_tensor      = */ ggml_backend_vk_buffer_cpy_tensor,

ggml/src/vulkan-shaders/argsort.comp

@@ -29,20 +29,18 @@ void main() {
     const int col = int(gl_LocalInvocationID.x);
     const uint row = gl_WorkGroupID.y;
 
-    if (col >= p.ncols_pad) {
-        return;
-    }
-
     const uint row_offset = row * p.ncols;
 
     // initialize indices
+    if (col < p.ncols_pad) {
         dst_row[col] = col;
+    }
     barrier();
 
     for (uint k = 2; k <= p.ncols_pad; k *= 2) {
         for (uint j = k / 2; j > 0; j /= 2) {
             const uint ixj = col ^ j;
-            if (ixj > col) {
+            if (col < p.ncols_pad && ixj > col) {
                 if ((col & k) == 0) {
                     if (dst_row[col] >= p.ncols ||
                         (dst_row[ixj] < p.ncols && (p.order == ASC ?

gguf-py/gguf/constants.py

@@ -94,6 +94,7 @@ class Keys:
         DECODER_START_TOKEN_ID            = "{arch}.decoder_start_token_id"
         ATTN_LOGIT_SOFTCAPPING            = "{arch}.attn_logit_softcapping"
         FINAL_LOGIT_SOFTCAPPING           = "{arch}.final_logit_softcapping"
+        SWIN_NORM                         = "{arch}.swin_norm"
         RESCALE_EVERY_N_LAYERS            = "{arch}.rescale_every_n_layers"
         TIME_MIX_EXTRA_DIM                = "{arch}.time_mix_extra_dim"
         TIME_DECAY_EXTRA_DIM              = "{arch}.time_decay_extra_dim"
@@ -237,6 +238,8 @@ class MODEL_ARCH(IntEnum):
     NEMOTRON     = auto()
     EXAONE       = auto()
     GRANITE      = auto()
+    GRANITE_MOE  = auto()
+    CHAMELEON    = auto()
 
 
 class MODEL_TENSOR(IntEnum):
@@ -345,6 +348,8 @@ class MODEL_TENSOR(IntEnum):
     ENC_FFN_DOWN         = auto()
     ENC_FFN_UP           = auto()
     ENC_OUTPUT_NORM      = auto()
+    CLS                  = auto() # classifier
+    CLS_OUT              = auto() # classifier output projection
 
 
 MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
@@ -396,6 +401,8 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
     MODEL_ARCH.NEMOTRON:       "nemotron",
     MODEL_ARCH.EXAONE:         "exaone",
     MODEL_ARCH.GRANITE:        "granite",
+    MODEL_ARCH.GRANITE_MOE:    "granitemoe",
+    MODEL_ARCH.CHAMELEON:      "chameleon",
 }
 
 TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
@@ -504,6 +511,8 @@ TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
     MODEL_TENSOR.ENC_FFN_DOWN:              "enc.blk.{bid}.ffn_down",
     MODEL_TENSOR.ENC_FFN_UP:                "enc.blk.{bid}.ffn_up",
     MODEL_TENSOR.ENC_OUTPUT_NORM:           "enc.output_norm",
+    MODEL_TENSOR.CLS:                       "cls",
+    MODEL_TENSOR.CLS_OUT:                   "cls.output",
 }
 
 MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
@@ -613,6 +622,8 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.FFN_DOWN,
         MODEL_TENSOR.FFN_UP,
         MODEL_TENSOR.LAYER_OUT_NORM,
+        MODEL_TENSOR.CLS,
+        MODEL_TENSOR.CLS_OUT,
     ],
     MODEL_ARCH.NOMIC_BERT: [
         MODEL_TENSOR.TOKEN_EMBD,
@@ -644,6 +655,7 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.FFN_GATE,
         MODEL_TENSOR.FFN_DOWN,
         MODEL_TENSOR.LAYER_OUT_NORM,
+        MODEL_TENSOR.CLS,
     ],
     MODEL_ARCH.MPT: [
         MODEL_TENSOR.TOKEN_EMBD,
@@ -807,6 +819,8 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.TOKEN_EMBD,
         MODEL_TENSOR.OUTPUT_NORM,
         MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ROPE_FACTORS_LONG,
+        MODEL_TENSOR.ROPE_FACTORS_SHORT,
         MODEL_TENSOR.ATTN_NORM,
         MODEL_TENSOR.ATTN_QKV,
         MODEL_TENSOR.ATTN_Q,
@@ -885,6 +899,8 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.TOKEN_EMBD,
         MODEL_TENSOR.OUTPUT_NORM,
         MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ROPE_FACTORS_LONG,
+        MODEL_TENSOR.ROPE_FACTORS_SHORT,
         MODEL_TENSOR.ATTN_NORM,
         MODEL_TENSOR.ATTN_Q_A,
         MODEL_TENSOR.ATTN_Q_B,
@@ -1250,6 +1266,7 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
     MODEL_ARCH.GRANITE: [
         MODEL_TENSOR.TOKEN_EMBD,
         MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
         MODEL_TENSOR.ATTN_NORM,
         MODEL_TENSOR.ATTN_Q,
         MODEL_TENSOR.ATTN_K,
@@ -1260,6 +1277,37 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.FFN_DOWN,
         MODEL_TENSOR.FFN_UP,
     ],
+    MODEL_ARCH.GRANITE_MOE: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.ATTN_Q,
+        MODEL_TENSOR.ATTN_K,
+        MODEL_TENSOR.ATTN_V,
+        MODEL_TENSOR.ATTN_OUT,
+        MODEL_TENSOR.FFN_NORM,
+        MODEL_TENSOR.FFN_GATE_INP,
+        MODEL_TENSOR.FFN_GATE_EXP,
+        MODEL_TENSOR.FFN_DOWN_EXP,
+        MODEL_TENSOR.FFN_UP_EXP,
+    ],
+    MODEL_ARCH.CHAMELEON: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.ATTN_Q,
+        MODEL_TENSOR.ATTN_Q_NORM,
+        MODEL_TENSOR.ATTN_K,
+        MODEL_TENSOR.ATTN_K_NORM,
+        MODEL_TENSOR.ATTN_V,
+        MODEL_TENSOR.ATTN_OUT,
+        MODEL_TENSOR.FFN_NORM,
+        MODEL_TENSOR.FFN_GATE,
+        MODEL_TENSOR.FFN_DOWN,
+        MODEL_TENSOR.FFN_UP,
+    ],
     # TODO
 }
 

gguf-py/gguf/gguf_writer.py

@@ -670,6 +670,9 @@ class GGUFWriter:
     def add_expert_weights_scale(self, value: float) -> None:
         self.add_float32(Keys.LLM.EXPERT_WEIGHTS_SCALE.format(arch=self.arch), value)
 
+    def add_swin_norm(self, value: bool) -> None:
+        self.add_bool(Keys.LLM.SWIN_NORM.format(arch=self.arch), value)
+
     def add_rescale_every_n_layers(self, count: int) -> None:
         self.add_uint32(Keys.LLM.RESCALE_EVERY_N_LAYERS.format(arch=self.arch), count)
 

gguf-py/gguf/tensor_mapping.py

@@ -87,6 +87,9 @@ class TensorNameMap:
             "rope.freqs",  # llama-pth
             "rotary_pos_emb.inv_freq",  # chatglm
         ),
+
+        MODEL_TENSOR.ROPE_FACTORS_LONG: (),
+        MODEL_TENSOR.ROPE_FACTORS_SHORT: (),
     }
 
     block_mappings_cfg: dict[MODEL_TENSOR, tuple[str, ...]] = {
@@ -256,6 +259,7 @@ class TensorNameMap:
             "model.layers.{bid}.mlp.gate",                      # qwen2moe olmoe
             "transformer.decoder_layer.{bid}.router",           # Grok
             "transformer.blocks.{bid}.ffn.router.layer",        # dbrx
+            "model.layers.{bid}.block_sparse_moe.router.layer", # granitemoe
         ),
 
         MODEL_TENSOR.FFN_GATE_INP_SHEXP: (
@@ -368,6 +372,7 @@ class TensorNameMap:
             "transformer.decoder_layer.{bid}.moe.linear_1",      # Grok (merged)
             "transformer.blocks.{bid}.ffn.experts.mlp.w2",       # dbrx
             "model.layers.{bid}.mlp.experts.down_proj",          # qwen2moe olmoe (merged)
+            "model.layers.{bid}.block_sparse_moe.output_linear", # granitemoe
         ),
 
         MODEL_TENSOR.FFN_DOWN_SHEXP: (
@@ -378,7 +383,7 @@ class TensorNameMap:
         MODEL_TENSOR.ATTN_Q_NORM: (
             "language_model.encoder.layers.{bid}.self_attention.q_layernorm",
             "model.layers.{bid}.self_attn.q_layernorm",                       # persimmon
-            "model.layers.{bid}.self_attn.q_norm",                            # cohere olmoe
+            "model.layers.{bid}.self_attn.q_norm",                            # cohere olmoe chameleon
             "transformer.blocks.{bid}.attn.q_ln",                             # sea-lion
             "encoder.layer.{bid}.attention.self.layer_norm_q",                # jina-bert-v2
             "transformer.layers.{bid}.attn.q_norm",                           # openelm
@@ -387,7 +392,7 @@ class TensorNameMap:
         MODEL_TENSOR.ATTN_K_NORM: (
             "language_model.encoder.layers.{bid}.self_attention.k_layernorm",
             "model.layers.{bid}.self_attn.k_layernorm",                       # persimmon
-            "model.layers.{bid}.self_attn.k_norm",                            # cohere olmoe
+            "model.layers.{bid}.self_attn.k_norm",                            # cohere olmoe chameleon
             "transformer.blocks.{bid}.attn.k_ln",                             # sea-lion
             "encoder.layer.{bid}.attention.self.layer_norm_k",                # jina-bert-v2
             "transformer.layers.{bid}.attn.k_norm",                           # openelm
@@ -681,6 +686,15 @@ class TensorNameMap:
         MODEL_TENSOR.ENC_OUTPUT_NORM: (
             "encoder.final_layer_norm", # t5
         ),
+
+        MODEL_TENSOR.CLS: (
+            "classifier",       # jina
+            "classifier.dense", # roberta
+        ),
+
+        MODEL_TENSOR.CLS_OUT: (
+            "classifier.out_proj", # roberta
+        ),
     }
 
     # architecture-specific block mappings

include/llama.h

@@ -102,6 +102,7 @@ extern "C" {
         LLAMA_VOCAB_PRE_TYPE_BLOOM          = 23,
         LLAMA_VOCAB_PRE_TYPE_GPT3_FINNISH   = 24,
         LLAMA_VOCAB_PRE_TYPE_EXAONE         = 25,
+        LLAMA_VOCAB_PRE_TYPE_CHAMELEON      = 26,
     };
 
     enum llama_rope_type {
@@ -192,6 +193,7 @@ extern "C" {
         LLAMA_POOLING_TYPE_MEAN = 1,
         LLAMA_POOLING_TYPE_CLS  = 2,
         LLAMA_POOLING_TYPE_LAST = 3,
+        LLAMA_POOLING_TYPE_RANK = 4, // used by reranking models to attach the classification head to the graph
     };
 
     enum llama_attention_type {
@@ -871,7 +873,8 @@ extern "C" {
 
     // Get the embeddings for a sequence id
     // Returns NULL if pooling_type is LLAMA_POOLING_TYPE_NONE
-    // shape: [n_embd] (1-dimensional)
+    // when pooling_type == LLAMA_POOLING_TYPE_RANK, returns float[1] with the rank of the sequence
+    // otherwise: float[n_embd] (1-dimensional)
     LLAMA_API float * llama_get_embeddings_seq(struct llama_context * ctx, llama_seq_id seq_id);
 
     //
@@ -910,6 +913,8 @@ extern "C" {
     //
     // Tokenization
     //
+    // The API is thread-safe.
+    //
 
     /// @details Convert the provided text into tokens.
     /// @param tokens The tokens pointer must be large enough to hold the resulting tokens.
@@ -1066,6 +1071,7 @@ extern "C" {
     LLAMA_API struct llama_sampler * llama_sampler_init_dist       (uint32_t seed);
 
     /// @details Sorts candidate tokens by their logits in descending order and calculate probabilities based on logits.
+    /// NOTE: Avoid using on the full vocabulary as the sorting can become slow. For example, apply top-k or top-p sampling first.
     LLAMA_API struct llama_sampler * llama_sampler_init_softmax    (void);
 
     /// @details Top-K sampling described in academic paper "The Curious Case of Neural Text Degeneration" https://arxiv.org/abs/1904.09751

models/ggml-vocab-chameleon.gguf.inp

@@ -0,0 +1,112 @@
+ied 4 ½ months
+__ggml_vocab_test__
+Führer
+__ggml_vocab_test__
+
+__ggml_vocab_test__
+ 
+__ggml_vocab_test__
+  
+__ggml_vocab_test__
+   
+__ggml_vocab_test__
+	
+__ggml_vocab_test__
+
+
+__ggml_vocab_test__
+
+
+
+__ggml_vocab_test__
+
+
+
+
+__ggml_vocab_test__
+	
+
+__ggml_vocab_test__
+Hello world
+__ggml_vocab_test__
+ Hello world
+__ggml_vocab_test__
+Hello World
+__ggml_vocab_test__
+ Hello World
+__ggml_vocab_test__
+ Hello World!
+__ggml_vocab_test__
+Hello, world!
+__ggml_vocab_test__
+ Hello, world!
+__ggml_vocab_test__
+ this is 🦙.cpp
+__ggml_vocab_test__
+w048 7tuijk dsdfhu
+__ggml_vocab_test__
+нещо на Български
+__ggml_vocab_test__
+កាន់តែពិសេសអាចខលចេញ
+__ggml_vocab_test__
+🚀 (normal) 😶‍🌫️ (multiple emojis concatenated) ✅ (only emoji that has its own token)
+__ggml_vocab_test__
+Hello
+__ggml_vocab_test__
+ Hello
+__ggml_vocab_test__
+  Hello
+__ggml_vocab_test__
+   Hello
+__ggml_vocab_test__
+    Hello
+__ggml_vocab_test__
+    Hello
+    Hello
+__ggml_vocab_test__
+ (
+__ggml_vocab_test__
+
+ =
+__ggml_vocab_test__
+' era
+__ggml_vocab_test__
+Hello, y'all! How are you 😁 ?我想在apple工作1314151天～
+__ggml_vocab_test__
+!!!!!!
+__ggml_vocab_test__
+3
+__ggml_vocab_test__
+33
+__ggml_vocab_test__
+333
+__ggml_vocab_test__
+3333
+__ggml_vocab_test__
+33333
+__ggml_vocab_test__
+333333
+__ggml_vocab_test__
+3333333
+__ggml_vocab_test__
+33333333
+__ggml_vocab_test__
+333333333
+__ggml_vocab_test__
+Cửa Việt
+__ggml_vocab_test__
+ discards
+__ggml_vocab_test__
+
+ 
+
+ 
+
+
+ 	 		 	
+  
+   
+    
+     
+🚀 (normal) 😶‍🌫️ (multiple emojis concatenated) ✅ 🦙🦙 3 33 333 3333 33333 333333 3333333 33333333 3.3 3..3 3...3 កាន់តែពិសេសអាច😁 ?我想在apple工作1314151天～ ------======= нещо на Български ''''''```````""""......!!!!!!?????? I've been 'told he's there, 'RE you sure? 'M not sure I'll make it, 'D you like some tea? We'Ve a'lL
+__ggml_vocab_test__

models/ggml-vocab-chameleon.gguf.out

@@ -0,0 +1,46 @@
+ 17245 16604 16403 16604 33583 18355
+ 16421 51153
+
+ 16604
+ 16650
+ 16650 16604
+ 16581
+ 16582
+ 16582 16582
+ 16582 16582 16582
+ 16581 16582
+ 31596 17394
+ 34926 17394
+ 31596 18671
+ 34926 18671
+ 34926 18671 16384
+ 31596 16395 17394 16384
+ 34926 16395 17394 16384
+ 16811 16704 20410 16483 16631 16397 52854
+ 16470 16399 16403 16407 16604 16406 35764 38185 51595 22592 26639
+ 29479 23955 17012 20103 25527 27670 17408 19005 21473 24774
+ 54254 42231 48084 29409 16617 61889 29409 16608 21954 16628 21954 16499 58445 29409 16607 58445 21954 16479 42231 21954 16611 21954 16607 21954 16633 21954 16611 29409 16607 21954 16615
+ 52351 16604 16391 25825 16392 23686 16498 39161 18885 16618 16488 30853 16604 16391 54124 17153 25134 16656 18476 26169 16895 16392 62193 16611 16604 16391 24664 17153 57169 16721 16872 17073 17304 28729 16392
+ 31596
+ 34926
+ 16650 31596
+ 16650 34926
+ 16696 31596
+ 16696 31596 16582 16696 31596
+ 16604 16391
+ 16582 16604 16412
+ 16390 22623
+ 31596 16395 16712 16390 16828 16384 17674 16769 16732 23686 16607 16604 16414 24427 16623 41809 16495 28999 36469 45292 30197 16400 16402 16400 16403 16400 16404 16400 43969 65211 16636
+ 16384 16384 16384 16384 16384 16384
+ 16402
+ 16402 16402
+ 16402 16402 16402
+ 16402 16402 16402 16402
+ 16402 16402 16402 16402 16402
+ 16402 16402 16402 16402 16402 16402
+ 16402 16402 16402 16402 16402 16402 16402
+ 16402 16402 16402 16402 16402 16402 16402 16402
+ 16402 16402 16402 16402 16402 16402 16402 16402 16402
+ 16418 19038 16639 16448 24315 33727 16467
+ 18765 17981
+ 16582 16604 16582 16582 16604 16582 16582 16582 16604 16581 16604 16581 16581 16604 16581 16582 16650 16582 16650 16604 16582 16696 16582 16696 16604 16582 52351 16604 16391 25825 16392 23686 16498 39161 18885 16618 16488 30853 16604 16391 54124 17153 25134 16656 18476 26169 16895 16392 62193 16611 20410 16483 16631 18885 16483 16631 16604 16402 16604 16402 16402 16604 16402 16402 16402 16604 16402 16402 16402 16402 16604 16402 16402 16402 16402 16402 16604 16402 16402 16402 16402 16402 16402 16604 16402 16402 16402 16402 16402 16402 16402 16604 16402 16402 16402 16402 16402 16402 16402 16402 16604 16402 16397 16402 16604 16402 16397 16397 16402 16604 16402 16397 16397 16397 16402 16604 54254 42231 48084 29409 16617 61889 29409 16608 21954 16628 21954 16499 58445 29409 16607 58445 21954 16479 42231 21954 16611 27683 16607 16604 16414 24427 16623 41809 16495 28999 36469 45292 30197 16400 16402 16400 16403 16400 16404 16400 43969 65211 16636 16604 16396 16396 16396 16396 16396 16396 16412 16412 16412 16412 16412 16412 16412 27268 23955 17012 20103 25527 27670 17408 19005 21473 24774 16604 16390 16390 16390 16390 16390 16390 16447 16447 16447 16447 16447 16447 16447 16385 16385 16385 16385 16397 16397 16397 16397 16397 16397 16384 16384 16384 16384 16384 16384 16414 16414 16414 16414 16414 16414 16687 16390 16690 16992 16604 16390 61797 16733 16390 16466 16986 16395 16604 16390 17879 16732 17811 16414 16604 16390 16428 16804 17811 16687 16390 16683 17190 16728 16395 16604 16390 16419 16732 16945 16991 25251 16414 17119 16390 38127 16641 16390 16459 16427

pyrightconfig.json

@@ -5,7 +5,8 @@
   "reportUnusedImport": "warning",
   "reportDuplicateImport": "error",
   "reportDeprecated": "warning",
-  "reportUnnecessaryTypeIgnoreComment": "warning",
+  "reportUnnecessaryTypeIgnoreComment": "information",
+  "disableBytesTypePromotions": false, // TODO: change once Python 3.12 is the minimum
   "executionEnvironments": [
     {
       // TODO: make this version override work correctly

requirements/requirements-convert_legacy_llama.txt

@@ -1,5 +1,5 @@
 numpy~=1.26.4
 sentencepiece~=0.2.0
-transformers>=4.40.1,<5.0.0
+transformers>=4.45.1,<5.0.0
 gguf>=0.1.0
 protobuf>=4.21.0,<5.0.0

scripts/compare-commits.sh

@@ -8,6 +8,9 @@ fi
 set -e
 set -x
 
+# verify at the start that the compare script has all the necessary dependencies installed
+./scripts/compare-llama-bench.py --check
+
 bench_args="${@:3}"
 
 rm -f llama-bench.sqlite > /dev/null

scripts/compare-llama-bench.py

@@ -92,6 +92,7 @@ help_s = (
     "If the columns are manually specified, then the results for each unique combination of the "
     "specified values are averaged WITHOUT weighing by the --repetitions parameter of llama-bench."
 )
+parser.add_argument("--check", action="store_true", help="check if all required Python libraries are installed")
 parser.add_argument("-s", "--show", help=help_s)
 parser.add_argument("--verbose", action="store_true", help="increase output verbosity")
 
@@ -99,6 +100,10 @@ known_args, unknown_args = parser.parse_known_args()
 
 logging.basicConfig(level=logging.DEBUG if known_args.verbose else logging.INFO)
 
+if known_args.check:
+    # Check if all required Python libraries are installed. Would have failed earlier if not.
+    sys.exit(0)
+
 if unknown_args:
     logger.error(f"Received unknown args: {unknown_args}.\n")
     parser.print_help()

scripts/sync-ggml.last

@@ -1 +1 @@
-10e83a412717c20d57ba19f025248e18e43addf3
+4de6ee8e6a4b2145d6b92162bc87722fecb4ea46

src/llama-impl.h

@@ -28,6 +28,8 @@ void llama_log_callback_default(ggml_log_level level, const char * text, void *
 #define LLAMA_LOG_INFO(...)  llama_log_internal(GGML_LOG_LEVEL_INFO , __VA_ARGS__)
 #define LLAMA_LOG_WARN(...)  llama_log_internal(GGML_LOG_LEVEL_WARN , __VA_ARGS__)
 #define LLAMA_LOG_ERROR(...) llama_log_internal(GGML_LOG_LEVEL_ERROR, __VA_ARGS__)
+#define LLAMA_LOG_DEBUG(...) llama_log_internal(GGML_LOG_LEVEL_DEBUG, __VA_ARGS__)
+#define LLAMA_LOG_CONT(...)  llama_log_internal(GGML_LOG_LEVEL_CONT , __VA_ARGS__)
 
 //
 // helpers

src/llama-sampling.cpp

@@ -3,13 +3,14 @@
 #include "llama-vocab.h"
 #include "llama-grammar.h"
 
-#include <cassert>
 #include <algorithm>
-#include <cstring>
+#include <cassert>
-#include <ctime>
 #include <cfloat>
 #include <chrono>
 #include <cmath>
+#include <cstdlib>
+#include <cstring>
+#include <ctime>
 #include <numeric>
 #include <random>
 #include <unordered_map>

src/llama-vocab.cpp

@@ -50,7 +50,7 @@ struct naive_trie {
             res.first->second.insert(key + 1, len - 1, value);
         }
     }
-    std::pair<const char *, size_t> get_longest_prefix(const char * key, size_t len, size_t offset = 0) {
+    std::pair<const char *, size_t> get_longest_prefix(const char * key, size_t len, size_t offset = 0) const {
         if (len == 0 || offset == len) {
             return std::make_pair(key, offset);
         }
@@ -79,6 +79,15 @@ struct naive_trie {
 // impl
 //
 
+struct llm_tokenizer {
+   llm_tokenizer() {}
+   virtual ~llm_tokenizer() = default;
+};
+
+llama_vocab::~llama_vocab() {
+    delete tokenizer;
+}
+
 int llama_vocab::find_bpe_rank(const std::string & token_left, const std::string & token_right) const {
     GGML_ASSERT(token_left.find(' ')   == std::string::npos);
     GGML_ASSERT(token_left.find('\n')  == std::string::npos);
@@ -187,10 +196,15 @@ struct llm_bigram_spm {
     size_t size;
 };
 
-struct llm_tokenizer_spm {
+struct llm_tokenizer_spm : llm_tokenizer {
-    llm_tokenizer_spm(const llama_vocab & vocab) : vocab(vocab) {}
+    llm_tokenizer_spm(const llama_vocab & /*vocab*/) : llm_tokenizer() {}
+};
+
+struct llm_tokenizer_spm_session {
+    llm_tokenizer_spm_session(const llama_vocab & vocab) : vocab(vocab) {}
 
     void tokenize(const std::string & text, std::vector<llama_vocab::id> & output) {
+
         // split string into utf8 chars
         int index = 0;
         size_t offs = 0;
@@ -279,7 +293,6 @@ private:
         if (left == -1 || right == -1) {
             return;
         }
-
         const std::string text = std::string(symbols[left].text, symbols[left].n + symbols[right].n);
         auto token = vocab.token_to_id.find(text);
 
@@ -306,10 +319,11 @@ private:
     }
 
     const llama_vocab & vocab;
+    // currently unused
+    // const llm_tokenizer_spm * spm_tokenizer;
 
     std::vector<llm_symbol> symbols;
     llm_bigram_spm::queue work_queue;
-
     std::map<std::string, std::pair<int, int>> rev_merge;
 };
 
@@ -352,8 +366,8 @@ struct llm_bigram_bpe {
     size_t size;
 };
 
-struct llm_tokenizer_bpe {
+struct llm_tokenizer_bpe : llm_tokenizer {
-    llm_tokenizer_bpe(const llama_vocab & vocab): vocab(vocab) {
+    llm_tokenizer_bpe(const llama_vocab & vocab) : llm_tokenizer() {
         GGML_ASSERT(vocab.type == LLAMA_VOCAB_TYPE_BPE);
         switch (vocab.type_pre) {
             case LLAMA_VOCAB_PRE_TYPE_LLAMA3:
@@ -450,6 +464,20 @@ struct llm_tokenizer_bpe {
                     "[^\\r\\n\\p{L}\\p{N}]?((?=[\\p{L}])([^a-z]))*((?=[\\p{L}])([^A-Z]))+|[^\\r\\n\\p{L}\\p{N}]?((?=[\\p{L}])([^a-z]))+((?=[\\p{L}])([^A-Z]))*|\\p{N}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n/]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
                 };
                 break;
+            case LLAMA_VOCAB_PRE_TYPE_CHAMELEON:
+                // Note: in theory, the special token (sentinel and image token) regex_exprs below
+                // are unnecessary, as they are split in `tokenizer_st_partition` anyway.
+                // However, since the upstream pre-tokenizer uses them, they are also
+                // included here (see https://huggingface.co/facebook/chameleon-7b).
+                regex_exprs = {
+                    "<sentinel:[0-9]+>",  // Sentinel tokens
+                    "(IMGIMG)((A|B|C|D|E|F|G|H|I){1,4})Z",  // Image tokens
+                    "([\\t\\n]|    |  )",  // directly from tokenizer.json
+                    "\\p{N}", // Individual digits
+                    "[\\p{P}!-/:-@\\[-`{-~]",  // Punctuation, Isolated
+                    "'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)",
+                };
+                break;
             default:
                 // default regex for BPE tokenization pre-processing
                 regex_exprs = {
@@ -462,7 +490,14 @@ struct llm_tokenizer_bpe {
         }
     }
 
-    void append(const llama_vocab::id token_id, std::vector<llama_vocab::id> & output) const {
+    std::vector<std::string> regex_exprs;
+};
+
+struct llm_tokenizer_bpe_session {
+    llm_tokenizer_bpe_session(const llama_vocab & vocab) : vocab(vocab),
+        bpe_tokenizer(static_cast<const llm_tokenizer_bpe *>(vocab.tokenizer)) {}
+
+    static void append(const llama_vocab::id token_id, std::vector<llama_vocab::id> & output)  {
         output.push_back(token_id);
     }
 
@@ -501,12 +536,11 @@ struct llm_tokenizer_bpe {
 
     void tokenize(const std::string & text, std::vector<llama_vocab::id> & output) {
         int final_prev_index = -1;
-
+        const auto word_collection = unicode_regex_split(text, bpe_tokenizer->regex_exprs);
-        const auto word_collection = unicode_regex_split(text, regex_exprs);
 
         symbols_final.clear();
 
-        for (auto & word : word_collection) {
+        for (const auto & word : word_collection) {
             work_queue = llm_bigram_bpe::queue();
             symbols.clear();
 
@@ -609,7 +643,6 @@ private:
         if (left == -1 || right == -1) {
             return;
         }
-
         std::string left_token  = std::string(symbols[left].text,  symbols[left].n);
         std::string right_token = std::string(symbols[right].text, symbols[right].n);
 
@@ -633,12 +666,10 @@ private:
     }
 
     const llama_vocab & vocab;
-
+    const llm_tokenizer_bpe * bpe_tokenizer;
-    std::vector<std::string> regex_exprs;
 
     std::vector<llm_symbol> symbols;
     std::vector<llm_symbol> symbols_final;
-
     llm_bigram_bpe::queue work_queue;
 };
 
@@ -646,15 +677,17 @@ private:
 // WPM tokenizer
 //
 
-struct llm_tokenizer_wpm {
+struct llm_tokenizer_wpm : llm_tokenizer {
-    llm_tokenizer_wpm(const llama_vocab & vocab): vocab(vocab) {}
+    llm_tokenizer_wpm(const llama_vocab & /*vocab*/) : llm_tokenizer() {}
+};
 
-    void tokenize(const std::string & text, std::vector<llama_vocab::id> & output) const {
+struct llm_tokenizer_wpm_session {
+    llm_tokenizer_wpm_session(const llama_vocab & vocab) : vocab(vocab) {}
+
+    void tokenize(const std::string & text, std::vector<llama_vocab::id> & output) {
         const auto & token_map = vocab.token_to_id;
-
         // normalize and split by whitespace
         std::vector<std::string> words = preprocess(text);
-
         // bos token prepended already
 
         // find the longest tokens that form the words
@@ -699,7 +732,7 @@ struct llm_tokenizer_wpm {
     }
 
     // TODO: reduce string copies by using cpts_offs array
-    std::vector<std::string> preprocess(const std::string & text) const {
+    static std::vector<std::string> preprocess(const std::string & text)  {
         const std::vector<uint32_t> cpts_nfd = unicode_cpts_normalize_nfd(unicode_cpts_from_utf8(text));
         std::vector<std::string> words(1, "");
 
@@ -751,15 +784,18 @@ struct llm_tokenizer_wpm {
             //(cpt >= 0xFF00  && cpt <= 0xFFEF);
     }
 
+private:
     const llama_vocab & vocab;
+    // currently unused
+    // const llm_tokenizer_wpm * wpm_tokenizer;
 };
 
 //
 // UGM tokenizer
 //
 
-struct llm_tokenizer_ugm {
+struct llm_tokenizer_ugm : llm_tokenizer {
-    llm_tokenizer_ugm(const llama_vocab & vocab) : vocab(vocab) {
+    llm_tokenizer_ugm(const llama_vocab & vocab) : llm_tokenizer() {
         if (vocab.precompiled_charsmap.size() > 0) {
             size_t charsmap_offset = 0;
 
@@ -805,6 +841,30 @@ struct llm_tokenizer_ugm {
         unknown_token_score = min_score - unknown_token_score_penalty;
     }
 
+    // escaped space symbol - U+2581 (Lower One Eighth Block)
+    const std::string escaped_space = "\xE2\x96\x81";
+
+    const char * prefix_replacements = NULL;
+    size_t prefix_replacements_size = 0;
+
+    const uint32_t * xcda_array = NULL;
+    size_t xcda_array_size = 0;
+
+    struct naive_trie user_defined_token_matcher;
+
+    float min_score = FLT_MAX;
+    float max_score = -FLT_MAX;
+
+    float unknown_token_score_penalty = 10.0;
+    float unknown_token_score;
+
+    struct naive_trie token_matcher;
+};
+
+struct llm_tokenizer_ugm_session {
+    llm_tokenizer_ugm_session(const llama_vocab & vocab) : vocab(vocab),
+        ugm_tokenizer(static_cast<const llm_tokenizer_ugm *>(vocab.tokenizer)) {}
+
     /* This implementation is based on SentencePiece optimized Viterbi algorithm for
      * unigram language models. The general idea is to:
      * - move along the input sequence in steps of one UTF code point,
@@ -843,7 +903,7 @@ struct llm_tokenizer_ugm {
             // traverse the token matcher trie to find a matching token
             bool single_codepoint_token_found = false;
             const struct best_tokenization & current_best = tokenization_results[input_offset];
-            const struct naive_trie * node = token_matcher.traverse(normalized[prefix_offset++]);
+            const struct naive_trie * node = ugm_tokenizer->token_matcher.traverse(normalized[prefix_offset++]);
 
             while (prefix_offset <= input_len && node != NULL) {
                 // check if we found valid token in prefix
@@ -873,7 +933,7 @@ struct llm_tokenizer_ugm {
             // if we didn't find a valid token corresponding to the whole UTF code point
             // then use unknown token as the tokenization of this UTF code point
             if (!single_codepoint_token_found) {
-                const double challenger_score = current_best.score_sum + unknown_token_score;
+                const double challenger_score = current_best.score_sum + ugm_tokenizer->unknown_token_score;
                 prefix_offset = input_offset + n_utf8_code_units;
                 struct best_tokenization & current_champ = tokenization_results[prefix_offset];
                 if (challenger_score > current_champ.score_sum) {
@@ -905,7 +965,6 @@ struct llm_tokenizer_ugm {
     }
 
 private:
-    const llama_vocab & vocab;
 
     // helper structure for returning normalization results
     struct normalization_result {
@@ -918,7 +977,7 @@ private:
         normalized->clear();
         normalized->reserve(input.size() * 3);
 
-        const std::string space = vocab.tokenizer_escape_whitespaces ? escaped_space : " ";
+        const std::string space = vocab.tokenizer_escape_whitespaces ? ugm_tokenizer->escaped_space : " ";
 
         bool shall_prepend_space = !vocab.tokenizer_treat_whitespace_as_suffix && vocab.tokenizer_add_space_prefix;
         bool shall_append_space = vocab.tokenizer_treat_whitespace_as_suffix && vocab.tokenizer_add_space_prefix;
@@ -1000,13 +1059,21 @@ private:
         size_t xcda_array_size;
     };
 
+    // this structure stores the best tokenization so far at input_offset
+    struct best_tokenization {
+        llama_token token_id;
+        size_t input_offset;
+        float score_sum;
+    };
+
     struct normalization_result normalize_prefix(const std::string & input, size_t input_offset) {
         if (input_offset == input.size()) {
             return { &input[input_offset], 0, 0 };
         }
 
         // if input prefix matches some user-defined token return this token as normalization result
-        auto user_defined_token_match = user_defined_token_matcher.get_longest_prefix(&input[input_offset], input.size() - input_offset);
+        auto user_defined_token_match =
+           ugm_tokenizer->user_defined_token_matcher.get_longest_prefix(&input[input_offset], input.size() - input_offset);
         if (user_defined_token_match.second > 0) {
             return { &input[input_offset], user_defined_token_match.second, user_defined_token_match.second };
         }
@@ -1014,8 +1081,8 @@ private:
         size_t longest_prefix_length = 0;
         size_t longest_prefix_offset = 0;
 
-        if (xcda_array_size > 0) {
+        if (ugm_tokenizer->xcda_array_size > 0) {
-            struct xcda_array_view xcda_view(xcda_array, xcda_array_size);
+            struct xcda_array_view xcda_view(ugm_tokenizer->xcda_array, ugm_tokenizer->xcda_array_size);
 
             // Find the longest normalized sequence matching the input prefix by walking
             // the XOR-compressed compact double array (XCDA) starting from the root node
@@ -1051,12 +1118,13 @@ private:
 
         if (longest_prefix_length > 0) {
             // we have a match, so return the replacement sequence
-            if (longest_prefix_offset >= prefix_replacements_size) {
+            if (longest_prefix_offset >= ugm_tokenizer->prefix_replacements_size) {
                 throw std::runtime_error("Index out of array bounds in precompiled charsmap!");
             }
-            const char * prefix_replacement = &prefix_replacements[longest_prefix_offset];
+            const char * prefix_replacement = &(ugm_tokenizer->prefix_replacements)[longest_prefix_offset];
             return { prefix_replacement, strlen(prefix_replacement), longest_prefix_length };
-        } else {
+        }
+
         // check if the input prefix contains a valid sequence of UTF-8 code units
         try {
             // if yes, return this sequence unmodified
@@ -1068,33 +1136,9 @@ private:
             return { "\xEF\xBF\xBD", 3, 1 };
         }
     }
-    }
 
-    // escaped space symbol - U+2581 (Lower One Eighth Block)
+    const llama_vocab & vocab;
-    const std::string escaped_space = "\xE2\x96\x81";
+    const llm_tokenizer_ugm * ugm_tokenizer;
-
-    const char * prefix_replacements = NULL;
-    size_t prefix_replacements_size = 0;
-
-    const uint32_t * xcda_array = NULL;
-    size_t xcda_array_size = 0;
-
-    struct naive_trie user_defined_token_matcher;
-
-    // this structure stores the best tokenization so far at input_offset
-    struct best_tokenization {
-        llama_token token_id;
-        size_t input_offset;
-        float score_sum;
-    };
-
-    float min_score = FLT_MAX;
-    float max_score = -FLT_MAX;
-
-    float unknown_token_score_penalty = 10.0;
-    float unknown_token_score;
-
-    struct naive_trie token_matcher;
 };
 
 //
@@ -1155,8 +1199,8 @@ static std::vector<uint8_t> llama_unescape_rwkv_token(const std::string & escape
     return output;
 }
 
-struct llm_tokenizer_rwkv {
+struct llm_tokenizer_rwkv : llm_tokenizer {
-    llm_tokenizer_rwkv(const llama_vocab & vocab): vocab(vocab) {
+    llm_tokenizer_rwkv(const llama_vocab & vocab) : llm_tokenizer() {
         // RWKV supports arbitrary byte tokens, but the vocab struct only supports string tokens.
         // For now, we decode the vocab here into the lookup we'll use for tokenization.
 
@@ -1168,11 +1212,17 @@ struct llm_tokenizer_rwkv {
         }
     }
 
+    struct naive_trie token_matcher;
+};
+
+struct llm_tokenizer_rwkv_session {
+    llm_tokenizer_rwkv_session(const llama_vocab & vocab) : vocab(vocab),
+        rwkv_tokenizer(static_cast<const llm_tokenizer_rwkv &>(*vocab.tokenizer)) {}
+
     void tokenize(const std::string & text, std::vector<llama_vocab::id> & output) {
         uint32_t position = 0;
-
         while (position < text.size()) {
-            const struct naive_trie * node = token_matcher.traverse(text[position]);
+            const struct naive_trie * node = rwkv_tokenizer.token_matcher.traverse(text[position]);
             if (node == NULL) {
                 // no matching token found, add unknown token
                 output.push_back(vocab.special_unk_id);
@@ -1197,11 +1247,33 @@ struct llm_tokenizer_rwkv {
         }
     }
 
+private:
     const llama_vocab & vocab;
-
+    const llm_tokenizer_rwkv & rwkv_tokenizer;
-    struct naive_trie token_matcher;
 };
 
+void llama_vocab::init_tokenizer() {
+    switch (type) {
+        case LLAMA_VOCAB_TYPE_SPM:
+            tokenizer = new llm_tokenizer_spm(*this);
+            break;
+        case LLAMA_VOCAB_TYPE_BPE:
+            tokenizer = new llm_tokenizer_bpe(*this);
+            break;
+        case LLAMA_VOCAB_TYPE_WPM:
+            tokenizer = new llm_tokenizer_wpm(*this);
+            break;
+        case LLAMA_VOCAB_TYPE_UGM:
+            tokenizer = new llm_tokenizer_ugm(*this);
+            break;
+        case LLAMA_VOCAB_TYPE_RWKV:
+            tokenizer = new llm_tokenizer_rwkv(*this);
+            break;
+        default:
+            GGML_ABORT("unsupported vocab type");
+    }
+}
+
 //
 // (de-) tokenize
 //
@@ -1263,7 +1335,7 @@ static void tokenizer_st_partition(const llama_vocab & vocab, std::forward_list<
 
             // if a fragment is text ( not yet processed )
             if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_RAW_TEXT) {
-                auto & raw_text = fragment.raw_text;
+                const auto & raw_text = fragment.raw_text;
 
                 auto raw_text_base_offset = fragment.offset;
                 auto raw_text_base_length = fragment.length;
@@ -1362,7 +1434,13 @@ static void tokenizer_st_partition(const llama_vocab & vocab, std::forward_list<
     }
 }
 
-std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab & vocab, std::string raw_text, bool add_special, bool parse_special) {
+std::vector<llama_vocab::id> llama_tokenize_internal(
+        const llama_vocab & vocab,
+        std::string raw_text,
+        bool add_special,
+        bool parse_special) {
+    GGML_ASSERT(vocab.tokenizer && "Tokenizer not initialized. Call llama_vocab::init_tokenizer() first.");
+
     std::vector<llama_vocab::id> output;
     std::forward_list<fragment_buffer_variant> fragment_buffer;
 
@@ -1399,9 +1477,9 @@ std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab & vocab,
 #ifdef PRETOKENIZERDEBUG
                         LLAMA_LOG_WARN("TT: (%ld %ld %ld) '%s'\n", raw_text.length(), fragment.offset, fragment.length, raw_text.c_str());
 #endif
-                        llm_tokenizer_spm tokenizer(vocab);
                         llama_escape_whitespace(raw_text);
-                        tokenizer.tokenize(raw_text, output);
+                        llm_tokenizer_spm_session session(vocab);
+                        session.tokenize(raw_text, output);
                         is_prev_special = false;
                     } else { // if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_TOKEN)
                         output.push_back(fragment.token);
@@ -1423,10 +1501,11 @@ std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab & vocab,
             } break;
         case LLAMA_VOCAB_TYPE_BPE:
             {
-                llm_tokenizer_bpe tokenizer(vocab);
+                llm_tokenizer_bpe_session session(vocab);
-
+                // it calls some other methods that are not exist in llm_tokenizer,
+                // here just cast it to bpe tokenizer object
                 if (add_special) {
-                    tokenizer.append_bos(output);
+                    session.append_bos(output);
                 }
                 for (const auto & fragment : fragment_buffer) {
                     if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_RAW_TEXT) {
@@ -1435,15 +1514,15 @@ std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab & vocab,
 #ifdef PRETOKENIZERDEBUG
                         LLAMA_LOG_WARN("TT: (%ld %ld %ld) '%s'\n", raw_text.length(), fragment.offset, fragment.length, raw_text.c_str());
 #endif
-                        tokenizer.tokenize(raw_text, output);
+                        session.tokenize(raw_text, output);
                     } else { // if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_TOKEN)
-                        tokenizer.append(fragment.token, output);
+                        session.append(fragment.token, output);
                     }
                 }
 
                 if (add_special) {
-                    tokenizer.append_eos(output);
+                    session.append_eos(output);
-                    tokenizer.check_double_bos_eos(output);
+                    session.check_double_bos_eos(output);
                 }
             } break;
         case LLAMA_VOCAB_TYPE_WPM:
@@ -1453,7 +1532,7 @@ std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab & vocab,
                     output.push_back(vocab.special_cls_id);
                 }
 
-                llm_tokenizer_wpm tokenizer(vocab);
+                llm_tokenizer_wpm_session session(vocab);
 
                 for (const auto & fragment : fragment_buffer) {
                     if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_RAW_TEXT) {
@@ -1462,7 +1541,7 @@ std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab & vocab,
 #ifdef PRETOKENIZERDEBUG
                         LLAMA_LOG_WARN("TT: (%ld %ld %ld) '%s'\n", raw_text.length(), fragment.offset, fragment.length, raw_text.c_str());
 #endif
-                        tokenizer.tokenize(raw_text, output);
+                        session.tokenize(raw_text, output);
                     } else { // if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_TOKEN)
                         output.push_back(fragment.token);
                     }
@@ -1475,12 +1554,11 @@ std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab & vocab,
             } break;
         case LLAMA_VOCAB_TYPE_UGM:
             {
-                llm_tokenizer_ugm tokenizer(vocab);
+                if (add_special && vocab.tokenizer_add_bos) {
-
-                if (add_special && vocab.tokenizer_add_bos != 0) {
                     GGML_ASSERT(vocab.special_bos_id != -1);
                     output.push_back(vocab.special_bos_id);
                 }
+                llm_tokenizer_ugm_session session(vocab);
 
                 for (const auto & fragment : fragment_buffer) {
                     if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_RAW_TEXT) {
@@ -1488,26 +1566,27 @@ std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab & vocab,
 #ifdef PRETOKENIZERDEBUG
                         LLAMA_LOG_WARN("TT: (%ld %ld %ld) '%s'\n", raw_text.length(), fragment.offset, fragment.length, raw_text.c_str());
 #endif
-                        tokenizer.tokenize(raw_text, output);
+                        session.tokenize(raw_text, output);
                     } else { // if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_TOKEN)
                         output.push_back(fragment.token);
                     }
                 }
 
-                if (add_special && vocab.tokenizer_add_bos != 0 && output.size() >= 2 && output[1] == vocab.special_bos_id) {
+                if (add_special && vocab.tokenizer_add_bos && output.size() >= 2 && output[1] == vocab.special_bos_id) {
                     LLAMA_LOG_WARN(
                         "%s: Added a BOS token to the prompt as specified by the model but the prompt "
                         "also starts with a BOS token. So now the final prompt starts with 2 BOS tokens. "
                         "Are you sure this is what you want?\n", __FUNCTION__);
                 }
 
-                if (add_special && vocab.tokenizer_add_eos == 1) {
+                if (add_special && vocab.tokenizer_add_eos) {
                     GGML_ASSERT(vocab.special_eos_id != -1);
                     output.push_back(vocab.special_eos_id);
                 }
             } break;
         case LLAMA_VOCAB_TYPE_RWKV:
             {
+                llm_tokenizer_rwkv_session session(vocab);
                 for (const auto & fragment : fragment_buffer) {
                     if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_RAW_TEXT) {
                         auto raw_text = fragment.raw_text.substr(fragment.offset, fragment.length);
@@ -1516,8 +1595,7 @@ std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab & vocab,
                         LLAMA_LOG_WARN("TT: (%ld %ld %ld) '%s'\n", raw_text.length(), fragment.offset, fragment.length, raw_text.c_str());
 #endif
 
-                        llm_tokenizer_rwkv tokenizer(vocab);
+                        session.tokenize(raw_text, output);
-                        tokenizer.tokenize(raw_text, output);
                     } else { // if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_TOKEN)
                         output.push_back(fragment.token);
                     }
@@ -1570,11 +1648,7 @@ llama_token_attr llama_token_get_attr_impl(const struct llama_vocab & vocab, lla
 }
 
 bool llama_token_is_eog_impl(const struct llama_vocab & vocab, llama_token token) {
-    return token != -1 && (
+    return token != -1 && vocab.special_eog_ids.count(token) > 0;
-        token == llama_token_eos_impl(vocab) ||
-        token == llama_token_eot_impl(vocab) ||
-        token == llama_token_eom_impl(vocab)
-    );
 }
 
 bool llama_token_is_control_impl(const struct llama_vocab & vocab, llama_token token) {
@@ -1717,11 +1791,13 @@ int32_t llama_token_to_piece_impl(const struct llama_vocab & vocab, llama_token
                 // suppressing them like CONTROL tokens.
                 if (attr & (attr_special | LLAMA_TOKEN_ATTR_USER_DEFINED)) {
                     return _try_copy(token_text.data(), token_text.size());
-                } else if (attr & LLAMA_TOKEN_ATTR_NORMAL) {
+                }
+                if (attr & LLAMA_TOKEN_ATTR_NORMAL) {
                     std::string result = token_text;
                     llama_unescape_whitespace(result);
                     return _try_copy(result.data(), result.size());
-                } else if (attr & LLAMA_TOKEN_ATTR_BYTE) {
+                }
+                if (attr & LLAMA_TOKEN_ATTR_BYTE) {
                     char byte = (char) llama_token_to_byte(vocab, token);
                     return _try_copy((char*) &byte, 1);
                 }
@@ -1732,7 +1808,8 @@ int32_t llama_token_to_piece_impl(const struct llama_vocab & vocab, llama_token
                 // suppressing them like CONTROL tokens.
                 if (attr & (attr_special | LLAMA_TOKEN_ATTR_USER_DEFINED)) {
                     return _try_copy(token_text.data(), token_text.size());
-                } else if (attr & LLAMA_TOKEN_ATTR_NORMAL) {
+                }
+                if (attr & LLAMA_TOKEN_ATTR_NORMAL) {
                     std::string result = llama_decode_text(token_text);
                     return _try_copy(result.data(), result.size());
                 }
@@ -1765,6 +1842,8 @@ int32_t llama_detokenize_impl(
                          int32_t   text_len_max,
                             bool   remove_special,
                             bool   unparse_special) {
+    GGML_ASSERT(vocab.tokenizer && "Tokenizer not initialized. Call llama_vocab::init_tokenizer() first.");
+
     int32_t avail = text_len_max;
     int32_t total = 0;
 

src/llama-vocab.h

@@ -6,6 +6,9 @@
 #include <vector>
 #include <unordered_map>
 #include <map>
+#include <set>
+
+struct llm_tokenizer;
 
 struct llama_vocab {
     using id    = llama_token;
@@ -49,6 +52,9 @@ struct llama_vocab {
     id special_eot_id    = -1; // TODO: move above after "eos_id", and here add "file separator" token
     id special_eom_id    = -1;
 
+    // set of all tokens that cause "end of generation"
+    std::set<id> special_eog_ids;
+
     // tokenizer flags
     bool tokenizer_add_space_prefix           = false;
     bool tokenizer_add_bos                    = false;
@@ -61,7 +67,14 @@ struct llama_vocab {
 
     std::vector<char> precompiled_charsmap;
 
+    llm_tokenizer * tokenizer = nullptr;
+
+    llama_vocab() = default;
+    ~llama_vocab();
+
     int find_bpe_rank(const std::string & token_left, const std::string & token_right) const;
+
+    void init_tokenizer();
 };
 
 //