* Introduce bfloat16 support
Many models on Hugging Face (e.g. Mistral, TinyLLaMA) use bfloat16 as
their canonical floating point format.
┌sign
│
│ ┌exponent
│ │
│ │ ┌mantissa
│ │ │
│┌──┴───┐┌─┴───┐
0b0000000000000000 brain16
This encoding has the same number of exponent bits as float32. That
makes conversion relatively straightforward, even in the absence of
hardware support. For example, converting brain16 to binary32 means
simply shifting 16 bits to the left.
┌sign
│
│ ┌exponent
│ │
│ │ ┌mantissa
│ │ │
│┌──┴───┐┌─┴───────────────────┐
0b00000000000000000000000000000000 IEEE binary32
The issue is that converting bf16 to fp16 can result in information
loss. Only 13% of bf16 numbers can be precisely represented in fp16
which in practice ends up being 99.71% of Mistral 7b v0.2's weights
however there is currently no way other than fp32 to get the others
┌sign
│
│ ┌exponent
│ │
│ │ ┌mantissa
│ │ │
│┌─┴─┐┌─┴──────┐
0b0000000000000000 IEEE binary16
This change fixes that, by adding a bf16 data type to GGML. Support
for CPU inference has been implemented along with optimizations for
the AVX2, AVX512, and AVX512BF16 ISAs. Perplexity on Mistral 7b 0.2
improves somewhere around -0.0024 to -0.0046 compared to using fp16
* Remove GGML code that's not needed
* Minimize the GGML API surface area for BF16
* Remove bf16 luts
* Make the GGML header look nicer
* Fix documentation
* Apply ggerganov's fixes for test-backend-ops
* Add BF16 code for new ggml_validate_row_data() function
* convert.py: add python logging instead of print()
* convert.py: verbose flag takes priority over dump flag log suppression
* convert.py: named instance logging
* convert.py: use explicit logger id string
* convert.py: convert extra print() to named logger
* convert.py: sys.stderr.write --> logger.error
* *.py: Convert all python scripts to use logging module
* requirements.txt: remove extra line
* flake8: update flake8 ignore and exclude to match ci settings
* gh-actions: add flake8-no-print to flake8 lint step
* pre-commit: add flake8-no-print to flake8 and also update pre-commit version
* convert-hf-to-gguf.py: print() to logger conversion
* *.py: logging basiconfig refactor to use conditional expression
* *.py: removed commented out logging
* fixup! *.py: logging basiconfig refactor to use conditional expression
* constant.py: logger.error then exit should be a raise exception instead
* *.py: Convert logger error and sys.exit() into a raise exception (for atypical error)
* gguf-convert-endian.py: refactor convert_byteorder() to use tqdm progressbar
* verify-checksum-model.py: This is the result of the program, it should be printed to stdout.
* compare-llama-bench.py: add blank line for readability during missing repo response
* reader.py: read_gguf_file() use print() over logging
* convert.py: warning goes to stderr and won't hurt the dump output
* gguf-dump.py: dump_metadata() should print to stdout
* convert-hf-to-gguf.py: print --> logger.debug or ValueError()
* verify-checksum-models.py: use print() for printing table
* *.py: refactor logging.basicConfig()
* gguf-py/gguf/*.py: use __name__ as logger name
Since they will be imported and not run directly.
* python-lint.yml: use .flake8 file instead
* constants.py: logger no longer required
* convert-hf-to-gguf.py: add additional logging
* convert-hf-to-gguf.py: print() --> logger
* *.py: fix flake8 warnings
* revert changes to convert-hf-to-gguf.py for get_name()
* convert-hf-to-gguf-update.py: use triple quoted f-string instead
* *.py: accidentally corrected the wrong line
* *.py: add compilade warning suggestions and style fixes
* ggml : add ggml_flash_attn_ext API
* ggml : fix GQA support in ggml_flash_attn_ext
* ggml : online attention (CPU)
* metal : initial implementation
* metal : f16 precision
* metal : reduce branches
* metal : specialize for head size
* wip : 8 rows per simd group
* wip : 4 rows per simd group
* wip : template for rows per warp
* metal : parallelize across KV size
* metal : parallel reduce across heads
* metal : efficient flash_attn_f16 implementation
* metal : avoid redundant loads of the attention
* metal : scale and mask in matrix form
* metal : fix comment
* llama : avoid ggml_cast, use F32 query
* metal : add parallel reduce version (disabled)
* metal : move output into local memory + optimize
- the result from each simdgroup now stays in the registers
- significantly reduced SRAM usage
- more efficient skipping of -INF blocks
- avoid simdgroup barrier in hot loop
- add comments
* metal : add tests, fix scaling, support C > 32
* metal : improve precision
* ggml : fix f16 mad
* metal : minor
* metal : support Q > 8
* tests : add ATTN tests
* metal : disable buffer allocation logs
* tests : more
* metal : faster inner loop for C == 32
* metal : fix array initialization
* tests : ifdef
* ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext
* ggml : fix ggml_soft_max mask requirement
* cuda : fix soft_max to use correct mask size
* cuda : add flash_attn kernel (wip)
* metal : optimize softmax for C > 32
* metal : optimize softmax
* tests : minor fix
* cuda : avoid zeroing fragments
* tests : update dims
* cuda : fix __hisinf() result check
* cuda : avoid warp_reduce for smax
* cuda : use int instead of int64_t
Noticeably improves performance (thanks to Johannes)
* cuda : make loops use the same loop values
Thanks Johannes again for the tip
* cuda : unroll some of the loops
* cuda : avoid __hisinf branches
* cuda : use half2 in softmax
* cuda : switch to 1 warp for bs > 16
* cuda : speed-up reduce part of the kernel
* cuda : unroll Q*K^T loop
* cuda : fix -INF block check
* cuda : simplify softmax
* cuda : fix matrix names
* cuda : minor
* llama : adapt to F16 KQ_pos
* llama : adapt new models to F16 KQ_mask
* ggml : fix F16 store (ARM NEON)
* llama : fix type of KQ_mask and KQ_pos
* ggml : fix CPU soft_max
* tests : add hs=256
* cuda : fix build
* metal : improve perf via smaller int registers
* cuda : adapt soft_max to F16 mask and pos
* CUDA: faster FlashAttention, kernel for bs == 1
* 16 cols for Phi-2
* no vec for hs, no hs==256 ncols==32 for Volta
* adjust kernel selection logic
* 4 warps, 256 stride for all D
* no ncols == 64
* Multiple parallel blocks for batch size 1
* fix compile warnings
* fix excessive KQ_b loads
* fix cmake build
* fix KV cache padding, NaN from INFINITY (#6438)
* llama : flash_attn cparam + fix defrag
* server: support flash_attn param
* server: bench: enable flash_attn param
* CUDA: refactor host code, dyn. par. blocks
* fix flash_attn_vec_f16 race condition
* flush softmax exp below threshold to 0
* store temp KQ in registers
* Calculate KQ as FP32 if KQV has GGML_PREC_F32
* Add __hgt2_mask implementation for CUDA 11
* fix KQ FP32 precision fpr parallel_blocks > 1
* llama-bench : add -fa,--flash-attn arg
* metal : add BS=1 kernel for flash attention (#6508)
* metal : add BS=1 kernel for flash attention (wip)
* metal : support more than 1 warps
* metal : opts
* metal : opt
* metal : switch to parallel reduce
* metal : reduce registers
* metal : simplify
* metal : initial FA vec kernel
* metal : use F32 attention accumulators
* batched-bench : add fattn arg
* llama : simplify llama_build_kv_store
ggml-ci
* llama : adapt build_olmo to changes
* ggml : fix arm fp16 store on windows
* metal : clean-up
* metal : clean-up kernel code
* metal : minor
* tests : remove benchmarks
ggml-ci
* ggml : fix avx512 const correctness
ggml-ci
* ggml : fix soft_max with bias on CPU
ggml-ci
* common : print --flash-attn in help
* ggml : fix num dimensions in ggml_flash_attn_ext
* llama : force disable flash attention for incompatible models
* ggml : ggml_soft_max support F16/F32 mask/pos
ggml-ci
* cuda : uint -> uint32_t
* cuda : "constexpr dim3" -> "const dim3"
ggml-ci
* cuda : try to fix __hgt2_mask
ggml-ci
* ggml : add TODO's for F16/F32 mask/pos support in other backends
* llama : replace bool need_kq_pos with use_alibi
* llama : prep ALiBi support for BERT models
ggml-ci
* llama : fix n_batch requirements
ggml-ci
* cont
* server : add help for --flash-attn arg
* llama : disable FA for AMD
* tests : remove TMP_ATTN_BENCH
ggml-ci
* llama : support save/load state with FA enabled
ggml-ci
* ci : add CUDA save-load-state tests
ggml-ci
* llama : llama_kv_cache_clear zeroes data + fix save-load seq
ggml-ci
* llama : fix copy-paste errors, add TODO
* llama : disallow incompatible states
* llama : update llama_state_get_size after v_trans field
* metal : remove tmp log
* llama : add static reminder for llama_state_get_size
* metal : fix max nsg
ggml-ci
* ci : fix arg order
ggml-ci
---------
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
* Cleaning up integration tests to share code between tests and make it simpler to add new tests.
* Add tests around quantifiers to ensure both matching and non-matching compliance.
* Add slightly more complex grammar with quantifiers to test references with quantifiers.
* Fixing build when C++17 is not present.
* Separating test calls to give more helpful stack traces on failure. Adding verbose messages to give visibility for what is being tested.
* Adding quotes around strings to explicitly show whitespace
* Removing trailing whitespace.
* Implementing suggestions from @ochafik -- grammars and test strings now print and flush before tests to aid in debugging segfaults and whatnot.
* Cleaning up forgotten symbols. Modifying simple test to use test harness. Added comments for more verbose descriptions of what each test is accomplishing.
* Unicode symbol modifications to hopefully make log easier to parse visually.
* ggml : group all experts in a single ggml_mul_mat_id
cuda : improve mmid row copy
* cuda : fix bin bcast with non-cont src0
* test-backend-ops : only run all mul mat tests for base types
* llama : disable moe offloading with SYCL
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* support qwen2moe
* fix-review
* metal : support unary ops for nelements % 4 != 0
* metal : require contiguousness for float4 unary kernels
* metal : require contiguousness for float4 unary kernels (cont)
* fix-review
* names : for brevity "SHARED_EXP" -> "SHEXP"
* llama : reuse build_moe_ffn()
* llama : add model type name
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Add chat template for command-r model series
* Fix indentation
* Add chat template test for command-r models and update the implementation to trim whitespaces
* Remove debug print
* Added integration tests for GBNF parser to validate correctness of parsing, as well as correctness of string matching. Intended for use to pin behavior while working on performance improvements.
* Fixing whitespace errors and cleaning error message alert to be clearer.
* Removing hacky include to llama.cpp from grammar integration test now that needed functions are available via internal API.
* Comment cleanup.
* Reorganizing tests for readability.
* Cleaning up debug message to make a bit more sense.
* Add openchat chat template
* Add chat template test for openchat
* Add chat template for vicuna
* Add chat template for orca-vicuna
* Add EOS for vicuna templates
* Combine vicuna chat templates
* Add tests for openchat and vicuna chat templates
* Add chat template for alpaca
* Add separate template name for vicuna-orca
* Remove alpaca, match deepseek with jinja output
* Regenerate chat template test with add_generation_prompt
* Separate deepseek bos from system message
* Match openchat template with jinja output
* Remove BOS token from templates, unprefix openchat
* ggml : update mul_mat_id to use the same tensor for all the experts
* update cuda
* minor
* update metal
* update test-backend-ops
* fix cuda
* Update ggml-metal.m
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* update convert.py
* update convert-hf-to-gguf.py
* update convert.py for mixtral hf models
* Update convert-hf-to-gguf.py
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* cuda : support non-pow-2 number of experts
* allow quantize to work for split and merged experts models in the same way
* cleanup + disable mmap automatically with split tensors models
* update imatrix
* test-backend-ops : test qwen argsort
* update grok model loading
* llama : add merged experts tensors to the grok tensor map
* minor
* gguf : bump version
* fix quantizing of merged experts
* convert-hf-to-gguf.py : update grok (untested)
* make linter happy
* cuda/argsort : use shared memory instead of pool memory
* convert : fix grok tensor names
* metal : add support for non-pow-2 argsort
* llama : more loader cleanup, better error checking
* cuda : fix warning
* llama : still use mmap for loading old models, but copy the data to a host buffer
* add review note
* llama : remove ffn tensor counting + add sanity check
ggml-ci
* convert : fix handling of n_experts == None
ggml-ci
* imatrix : fix ncall counters
* llama : produce error if imatrix size does not match
* quantize : terminate on errors + trace logs
ggml-ci
* metal : pad shared memory to 16 bytes
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* iq1_m: basics
* iq1_m: basics-2
* iq1_m: CUDA dequantize works
Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B.
* iq1_m: separate shifts for each group of 8 in a block
We get
PPL(LLaMA-v2-7B ) = 9.2810
PPL(LLaMA-v2-13B) = 6.8105
Not bad, but slightly higher than
sqrt(PPL(IQ1_S) * PPL(IQ2_XXS))
which is the expected outcome given that IQ1_M is
halfway between IQ1_S and IQ2_XXS in terms of bpw.
From this, we would expect
PPL = 9.14 for LLaMA-v2-7B
PPL = 6.63 for LLaMA-v2-13B
* iq1_m: go to 3-bit scales
There is slight increase in PPL, but the 0.0625 bpw reduction
in size is totally worth it.
We now have
PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw
PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw
PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw
* iq1_m: scalar dot product
* iq1_m: AVX2 dot product
* iq1_m: very slightly faster AVX2 dot product
* iq1_m: ARM_NEON dot product
Works, but very slow (10.5 t/s)
* iq1_m: Metal - dequantize works, dot product does not
* iq1_m: Metal now works
About the same performance as iq1_s.
* iq1_m: minor
* iq1_m: checking pure iq1_m quantization
It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight
with Q4_K.
* iiq1_m: slightly faster ARM_NEON dot product
10.5 t/s -> 11.65 t/s
* iq1_m: faster ARM_NEON dot product
11.65 t/s -> 14.9 t/s
* iq1_m: another minor ARM_NEON dot product improvement
14.9 -> 15.0 t/s
* iq1_m: small PPL improvement via super-block scale adjustment
After quantizing block scales redo the super-block scale fit.
PPL(LLaMA-v2-7B ) = 9.3346
PPL(LLaMA-v2-13B) = 6.8419
PPL(LLaMA-v2-70B) = 4.8294
PPL(Mistral-7B ) = 8.1624
* iq1_m: adapt to CUDA refactoring
* iq1_m: remove unused variable
We have progressed to warnings being errors.
* iq1_m: add to backend-ops tests
* iq1_m: fix Windows ARM
* iq1_m: use common definition of iq1m_scale_t
* cuda: assert -> NO_DEVICE_CODE
* iq1_M: PR comments
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* Try IQ4_NL with blocks of 64 - does not look good
* iq4_xs: go to super-blocks of 256 and 6-bit scales for blocks of 32
* iq4_xs: CUDA works - 133.2 t/s
* iq4_xs: AVX2 dot product
* iq4_xs: ARM_NEON dot product
* iq4_nl: Metal implementation
As usual, Metal / Apple Silicon don't like my quants.
* iq3_xs: minor fix
* iq4_xs: shrink by using IQ3_S for attn_k and attn_q
* iq4_xs: revert using IQ3_S for attn_k and attn_v
PPL vs size is good, but CPU performance suffers: on M2 Max
TG-128 drops to 21.7 t/s from 28.8, and on a Ryzen-7950X
to 14.5 t/s from 15.8 t/s. On CUDA we have 135 t/s when
using IQ3_S vs 133 t/s with pure IQ4_XS.
* Fix CI
* iq4_xs: Added forgotten check for 256 divisibility
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* Adding IQ2_S and IQ2_M as a single cumulative commit
* Update examples/quantize/quantize.cpp
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* iq4_nl: squash commits for easier rebase
* Basics (quantize, dequantize)
* CUDA dequantize and dot product
* Slightly faster CUDA dot product (120 t/s)
* Switch to 6-bit scales
* Scalar dot product
* AVX2 dot product
* ARM_NEON dot product
* Works on metal, but still slow
* Slightly better Metal dot product
* Another small Metal improvement
* Metal dot product is getting there
* Faster CUDA dot product
* Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided
* Report the actual bpw
* Add _xs mix that is 4.05 bpw for non-MoE models
* Remove IQ4_XS for now, slightly adjust kvalues_iq4nl
* AVX2 dot product uses Q8_0 instead of Q8_K
* Add to test-backend-ops
* Minor fix
* Also use use Q5_K for attn_output in MoE models
* Fixes after merging latest master
* Switching to blocks of 32
* AVX2 for blocks of 32
* Scaler dot product for blocks of 32
* ARM_NEON dot product for blocks of 32
* Metal kernels for blocks of 32
* Slightly faster Metal kernels
* Resurrecting iq3_xs
After all the experimentation, nothing was better than this.
* Minor PPL improvement via a block scale fudge factor
* Minor improvement via 3 neighbours
* iq3_xs: working scalar and AVX2 dot products
* iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s)
* iq3_xs: working Metal implementation
* Adding IQ3_M - IQ3_XS mix with mostly Q4_K
* iiq3_xs: a 3.4375 bpw variant
* iq3_xs: make CUDA work for new version
* iq3_xs: make scalar and AVX2 work for new version
* iq3_s: make ARM_NEON work with new version
* iq3_xs: make new version work on metal
Performance is very similar to Q3_K_S
* iq3_xs: tiny Metal speed improvement
* iq3_xs: tiny Metal speed improvement
* Fix stupid warning
* Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS
* iq3_xs: rename to iq3_s
* iq3_s: make tests pass
* Move Q3_K_XS mix to 3.25 bpw
* Attempt to fix failing tests
* Another attempt to fix the Windows builds
* Attempt to fix ROCm
* ROCm again
* iq3_s: partial fix for QK_K = 64
* iq3_s: make it work on metal for QK_K = 64
Pleasent surprise: the coding was super-block size independent,
so all it took was to delete some QK_K == 256 guards.
* Will this fix ROCm?
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* server: fallback to chatml
* add new chat template
* server: add AlphaMonarch to test chat template
* server: only check model template if there is no custom tmpl
* remove TODO
* iq4_nl: squash commits for easier rebase
* Basics (quantize, dequantize)
* CUDA dequantize and dot product
* Slightly faster CUDA dot product (120 t/s)
* Switch to 6-bit scales
* Scalar dot product
* AVX2 dot product
* ARM_NEON dot product
* Works on metal, but still slow
* Slightly better Metal dot product
* Another small Metal improvement
* Metal dot product is getting there
* Faster CUDA dot product
* Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided
* Report the actual bpw
* Add _xs mix that is 4.05 bpw for non-MoE models
* Remove IQ4_XS for now, slightly adjust kvalues_iq4nl
* AVX2 dot product uses Q8_0 instead of Q8_K
* Add to test-backend-ops
* Minor fix
* Also use use Q5_K for attn_output in MoE models
* Fixes after merging latest master
* Switching to blocks of 32
* AVX2 for blocks of 32
* Scaler dot product for blocks of 32
* ARM_NEON dot product for blocks of 32
* Metal kernels for blocks of 32
* Slightly faster Metal kernels
* iq4_nl: Fix after merging with master
* iq4_nl: another fix after merging with master
* Use IQ4_NL instead of Q4_K when using k-quants is not possible
* Fix typo that makes several tests fail
* It was the ggml_vdotq thing missed inside the brackets
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* llama: add llama_chat_apply_template
* test-chat-template: remove dedundant vector
* chat_template: do not use std::string for buffer
* add clarification for llama_chat_apply_template
* llama_chat_apply_template: add zephyr template
* llama_chat_apply_template: correct docs
* llama_chat_apply_template: use term "chat" everywhere
* llama_chat_apply_template: change variable name to "tmpl"
* iq1_s: WIP basics
* iq1_s: CUDA is working
* iq1_s: scalar CPU dot product
* iq1_s: WIP AVX2 dot product - something is not right
* Fix tests
* Fix shadow warnings
* Fix after merge with latest master
* iq1_s: AVX2 finally works
* iq1_s: ARM_NEON dot product. Works, but not very fast
* iq1_s: better grid
* iq1_s: use IQ2_XXS for attn_output
At a cost of 0.04 extra bpw this gives a big improvement in PPL.
* iq1_s: Metal basics
Dequantize works, but not dot product
* iq1_s: Metal works, but quite slow
As usual, Apple Silicon does not like the code I write.
* iq1_s: Tests
* iq1_s: slightly faster dot product
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* ggml : avoid recomputing alibi slopes (CPU)
* llama : reuse hparams.f_max_alibi_bias in all cases
ggml-ci
* ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal)
ggml-ci
* ggml : handle all SRCs (do not break on first null)
ggml-ci
* tests : do not use slope for large soft_max
accumulates too much error
ggml-ci
* ggml : alternative ALiBi without extra tensor
We compute the slopes in the kernel
ggml-ci
* cuda : add ALiBi support in ggml_soft_max_ext
ggml-ci
* ggml : deprecate ggml_alibi
* ggml : support multi-sequence ALiBi (Metal)
ggml-ci
* cuda : add multi-seq ALiBi + remote F16 soft_max
ggml-ci
* ggml : update deprecation message
* ggml : fix pos ptr when no ALiBi
ggml-ci
* cuda : fix performance (pow -> powf)
* cuda : precompute ALiBi constants
* metal : pre-compute ALiBi slopes
ggml-ci
* llama : init kq_pos only if needed
ggml-ci
* test-backend-ops : add null pos test to soft_max
test-backend-ops : replace soft_max tests
ggml-ci
---------
Co-authored-by: slaren <slarengh@gmail.com>
