* Support for Yi-VL, templating fix for mobileVLM
* ws
* Update examples/llava/clip.cpp
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update llava-cli.cpp
* Update clip.cpp
bugfix for new conversions
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* server: add llama_server_queue struct
* server: add llama_server_response_event
* server: add comments
* server: move all mutexes away from server.cpp
* server: correct multitask response
* server: only add back deferred tasks when one slot is available
* server: fix a race condition cause by "request_completion"
* kl-divergence: be able to save all logits to a file
* Add ability to compute KL-divergence
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* MobileVLM native implementation
* delete depthwise_conv_2d and permute_cpy relative code, replace the two by the existed functions, and opt ldp definition, support LLAMA_PERF option for CMake
* move android script to example/llava directory
* Fix the editor config checks
---------
Co-authored-by: Chenxiaotao03 <chenxiaotao03@meituan.com>
This commit adds `--sample-start` and `--include-sample-start` to the
output from the main function in finetune.cpp.
The motivation for this is that even though these are set explicitly by
the user via the command line, if one forgets to set them then it is
useful to have their values printed out. Otherwise it is possible to go
through the whole training process before realizing that the values are
not what one expected.
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
* Add Q3_K_XS - intermediate size between Q2_K and Q3_K_S
* Q3_K_XS: quanize first 1/8 of ffn_down layers with Q4_K
Together with an importance matrix, this brings perplexity
for LLaMA-v2-70B below the perplexity of the former Q2_K
with a 800 MB smaller quantized model size.
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* TruthfulQA: 1st attempt, does not look like it is working
The same implementation can be used for HellaSwag as well,
so I converted a HellaSwag validation dataset to the binary
format used here and tested with that. The score is only
around 50, so something is not quite right.
* TruthfulQA: works but the result is bad
I know it works because if I convert the HellaSwag validation
data to the binary format used in the truthful_qa_score() function
I get the exact same result as from the hellaswag_score() function.
But I guess, the questions are tricky and the way I have done
the combination of question + answer is very likely not the best.
The TruthfulQA validation dataset contains 817 questions, with
random chance result around 19%. With this version I get
29.1% for Mistral-7B and 55.2% for Mistral-7B-Instruct-v0.2.
The HF leader board results for these two models are
42.2% and 68.3%, respectively.
* TruthfulQA: fix random sample
* TruthfulQA: prepare tasks in parallel for large test datasets
* Rename truthful_qa to multiple_choice
* Make MSVC happy
I had forgotten that MSVC does not make constexpr's available
inside a lambda.
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
For Mistral-7B and fp16, time on my system goes down from 536 seconds
to 423 seconds for the full evaluation dataset (10042 tasks).
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* winogrande: simple implementation
It doesn't look like it is working - why?
For Mistral-7B it is barely better than
random chance (score ~60% for 1267 tasks), while I see
Mistral-7B scoring 78.4% on the HF leader board.
1-sigma statistical uncertainty for 1267 tasks is ~1.4,
so no way the difference is due to statistics.
* winogrande: somewhat better
Score for Mistrali7-B is now 68.9 on the validation set of
winogrande_debiased. Still far from the reported 78.4, but
better than what I had before.
* winogrande: improving
Mistral-7B score is now 73.56.
Still not quite 78.4 but getting there.
We are also getting a lower score on HellaSwag
compared to HF leader board, so I'm not expecting
we will get up to 78.4 anyway.
It looks like it is better to skip the choice word(s)
when evaluating the average log-likelihood. This kind of
makes sense because a more common word (in Winogrande this is
often a name) will have a higher probability without knowing
about the follow up context, and this will skew the log-likelihood
towards the more common word. We can only do this if the
choice words are not last in the sentence.
It also looks like it is better to skip the punctuation at the
end of the sentence, provided the choice words are not last.
* winogrande: add dataset instructions
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
* backend : add eval callback
ggml-ci
* backend : group nodes in a single compute when user don't need them
* backend : clean-up the implementation
ggml-ci
* simple : do not perform tensor data copy if not needed
* simple : fix
* imatrix : offload to GPU support
* imatrix : fix ggml_mul_mat_id hanlding
ggml-ci
* ci : add imatrix test
ggml-ci
* ci : rearrange output
ggml-ci
This commit adds the name of the training data file to the log message
printed when the training data is tokenized.
The motivation for this change is that it can be useful to show which
file is being tokenized when running the finetune example.
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
* Introduce starter project for Android
Based on examples/llama.swiftui.
* Add github workflow
* Set NDK version
* Only build arm64-v8a in CI
* Sync bench code
* Rename CI prop to skip-armeabi-v7a
* Remove unused tests
This commit replaces the magic number LLAMA_FILE_MAGIC_LORA used in
finetune.cpp with LLAMA_FILE_MAGIC_GGLA defined in llama.h.
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
* examples : save-load-state: save only required state
* llama : only reserve n_vocab * n_batch at most for logits
llama_decode asserts that only n_batch tokens are passed each call, and
n_ctx is expected to be bigger than n_batch.
* llama : always reserve n_vocab * n_batch for logits
llama_context de-serialization breaks if the contexts have differing
capacity for logits and llama_decode will at maximum resize to
n_vocab * n_batch.
* llama : only save and restore used logits
for batch sizes of 512 this reduces save state in the best case by
around 62 MB, which can be a lot if planning to save on each message
to allow regenerating messages.
* llama : use ostringstream and istringstream for save and load
* llama : serialize rng into minimum amount of space required
* llama : break session version due to serialization changes
* add the parameter : --no-display-prompt , combine with --log-disable it will display only the generated tokens
* remove empty line
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* llama : ggml-backend integration
* ggml-backend : add names to buffers
* fix unmap after loading
* batched-bench : add tensor_split param
* llama : check for null tensor_split
* ggml-backend : increase GGML_MAX_BACKENDS
* improve graph splitting, partial fix for --no-kv-offload
* cuda : add ggml-backend split buffer support
* cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available)
* ggml : fix null backend dereference (#4807)
* ggml : fix null backend dereference
* ggml : also check ggml_backend_is_cpu
* test-backend-ops : check buffer allocation failures
* llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row)
* ggml : fix mul_mat_id work size
* llama : rewrite session kv load/set without graphs
* minor
* llama : only initialize used backends, free backends on context free
* llama : abort ctx if cuda backend init fails
* llama : rewrite lora with ggml-backend and compute on CPU
ggml-ci
* llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer
* opencl : add ggml-backend buffer type
* cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf)
* llama : on Metal, by default offload the full model
ggml-ci
* metal : page align the data ptr (#4854)
* Apply suggestions from code review
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* cuda : fix split buffer free
* address review comments
* llama-bench : add split-mode parameter
* fix whitespace
* opencl : fix double initialization
* server : add --split-mode parameter
* use async copy and compute to improve multi-gpu performance
ggml-ci
* use async memcpys to copy the graph outputs to the CPU
* fix opencl
* use a host buffer for the cpu compute buffer for faster copies to the gpu
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
This commit replaces the magic number used in export-lora.cpp with
the one defined in llama.h, which is indirectly included via common.h.
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
* Updated Models Layout
- Added a models drawer
- Added downloading directly from Hugging Face
- Load custom models from local folder
- Delete models by swiping left
* trimmed trailing white space
* Updated Models Layout
* Restore intended k-quants quantization mixes for MoE models
* Update Q2_K_S values in the quantize tool
Still using LLaMA-v1 PPL values in the quant description
today does not make much sense. But let's leave this update
for another PR.
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* server: added support for multiple api keys, added loading api keys from file
* minor: fix whitespace
* added file error handling to --api-key-file, changed code to better
reflect current style
* server: update README.md for --api-key-file
---------
Co-authored-by: Michael Coppola <info@michaeljcoppola.com>
* added /health endpoint to the server
* added comments on the additional /health endpoint
* Better handling of server state
When the model is being loaded, the server state is `LOADING_MODEL`. If model-loading fails, the server state becomes `ERROR`, otherwise it becomes `READY`. The `/health` endpoint provides more granular messages now according to the server_state value.
* initialized server_state
* fixed a typo
* starting http server before initializing the model
* Update server.cpp
* Update server.cpp
* fixes
* fixes
* fixes
* made ServerState atomic and turned two-line spaces into one-line
* updated `server` readme to document the `/health` endpoint too
* used LOG_INFO after successful model loading
* added /health endpoint to the server
* added comments on the additional /health endpoint
* Better handling of server state
When the model is being loaded, the server state is `LOADING_MODEL`. If model-loading fails, the server state becomes `ERROR`, otherwise it becomes `READY`. The `/health` endpoint provides more granular messages now according to the server_state value.
* initialized server_state
* fixed a typo
* starting http server before initializing the model
* Update server.cpp
* Update server.cpp
* fixes
* fixes
* fixes
* made ServerState atomic and turned two-line spaces into one-line
* updated `server` readme to document the `/health` endpoint too
* added /health endpoint to the server
* added comments on the additional /health endpoint
* Better handling of server state
When the model is being loaded, the server state is `LOADING_MODEL`. If model-loading fails, the server state becomes `ERROR`, otherwise it becomes `READY`. The `/health` endpoint provides more granular messages now according to the server_state value.
* initialized server_state
* fixed a typo
* starting http server before initializing the model
* Update server.cpp
* Update server.cpp
* fixes
* fixes
* fixes
* made ServerState atomic and turned two-line spaces into one-line
Uses ggml functions instead of hardcoded names and adds support to quantize into the modern Q-K variants.
This is just the bare minimum to get k-types working - a more refined choice of types would be needed to get best quality on low quantizations.
I ran a few tests, it doesn't break anything I could notice and a Q6_K ViT works almost as well as Q8_0 but 3 times the inference speed.
This change fixes an issue where supplying `--image missing-file` would
result in a segfault due to a null pointer being dereferenced. This can
result in distracting info being printed if robust crash analysis tools
are being used.
* updated server readme to reflect the gg/server-token-probs-4088 commit
added explanation for the API's completion result which now includes `completion_probabilities`. Also added a JSON schema that shows the type/structure of `completion_probabilities`.
* simplified the `completion_probabilities` JSON schema
It's now easier to understand what the structure of `completion_probabilities` looks like.
* minor : fix trailing whitespace
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Changes to server to allow metadata override
* documentation
* flake.nix: expose full scope in legacyPackages
* flake.nix: rocm not yet supported on aarch64, so hide the output
* flake.nix: expose checks
* workflows: nix-ci: init; build flake outputs
* workflows: nix-ci: add a job for eval
* workflows: weekly `nix flake update`
* workflows: nix-flakestry: drop tag filters
...and add a job for flakehub.com
* workflows: nix-ci: add a qemu job for jetsons
* flake.nix: suggest the binary caches
* flake.lock: update
to a commit recently cached by nixpkgs-cuda-ci
---------
Co-authored-by: John <john@jLap.lan>
Co-authored-by: Someone Serge <sergei.kozlukov@aalto.fi>
This change makes it possible to use flags like `--grammar` when using
the `llava-cli` program. The rest is just code cleanup deleting a long
standing TODO comment.
This change also ensures that logging information is emitted to stderr
which helps the `llava-cli` command be more friendly to shell scripts.
See Mozilla-Ocho/llamafile@1cd334f
The server currently schedules tasks using a sleep(5ms) busy loop. This
adds unnecessary latency since most sleep implementations do a round up
to the system scheduling quantum (usually 10ms). Other libc sleep impls
spin for smaller time intervals which results in the server's busy loop
consuming all available cpu. Having the explicit notify() / wait() code
also helps aid in the readability of the server code.
See mozilla-Ocho/llamafile@711344b
The default values for tfs_z and typical_p were being set to zero, which
caused the token candidates array to get shrunk down to one element thus
preventing any sampling. Note this only applies to OpenAI API compatible
HTTP server requests.
The solution is to use the default values that OpenAI documents, as well
as ensuring we use the llama.cpp defaults for the rest. I've tested this
change still ensures deterministic output by default. If a "temperature"
greater than 0 is explicitly passed, then output is unique each time. If
"seed" is specified in addition to "temperature" then the output becomes
deterministic once more.
See mozilla-Ocho/llamafile#117
See mozilla-Ocho/llamafile@9e4bf29
* initial commit, going through initializations
* main loop finished, starting to debug
* BUG: generates gibberish/repeating tokens after a while
* kv_cache management
* Added colors to distinguish drafted tokens (--color). Updated README
* lookup : fix token positions in the draft batch
* lookup : use n_draft from CLI params
* lookup : final touches
---------
Co-authored-by: Leon Ericsson <leon.ericsson@icloud.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Add API key authentication for enhanced server-client security
* server : to snake_case
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Fixes "Not enough space in the context's memory pool" encountered on certain models, which seems to be caused by some imprecision related to the automatic casting of floating point values
* do not cast to size_t, instead just use doubles
* ggml : add ggml_row_size(), deprecate ggml_type_sizef()
* ggml : fix row size compute to avoid overflows
* tests : fix sizey -> sizez
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
On commit b1108 (44c117f4) xaedes added
ggml_allocr * alloc = NULL;
... (many lines in between)
if (alloc) {
ggml_allocr_free(alloc);
}
Which is correct, but it's easy to lose context after many lines in between.
On commit b1287 (0e76a899) xaedes made a big change. From here on, alloc is freed eagerly.
alloc = ggml_allocr_new(...)
... (short lines of code)
ggml_allocr_free(alloc)
This happens a few times, but alloc is never set to NULL, and many lines below,
we still have
if (alloc) {
ggml_allocr_free(alloc);
}
which causes a double-free.
* Samplers sequence order w parameter
* Cleaned commented code
* Fixed formatting
* Rewrote with unordered_map
* Revert and rewrite, too many problems and safeguards would be needed
* Fixed code style
* Code style fixes according to review
* More readable samplers input string, fixed help
* Style fix in sampler_queue
* Formatting fixes
* Fixing whitespaces
This commit updates the error message that is printed when the
KV cache is not big enough to hold all the prompt and generated
tokens. Specifically it removes the reference to n_parallel and
replaces it with n_len.
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
* metal : implement soft_max_ext
* cuda : implement soft_max_ext
* ggml : implement soft_max_ext (CPU)
* batched-bench : print threads
ggml-ci
* metal : simplify soft_max encoding
ggml-ci
* cuda : use 512 threads for soft_max instead of 32
* ggml : update soft max cpu
* cuda : do warp-based block reduce
* cuda : increase max block size to 1024
* cuda : fix warp reduction initialization of shared mem
* metal : warp-based reduction for soft max kernel
* metal : warp-based reduce for rms_norm
* metal : simplify soft max kernel
ggml-ci
* alloc : fix build with debug
* * add multiprompt support
* * cleanup
* * more cleanup
* * remove atomicity of id_gen, and change lock_guard to unique_lock on completion requests
* * remove all references to mutex_multitasks
* Update examples/server/server.cpp
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* Update examples/server/server.cpp
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* Update examples/server/server.cpp
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* Update examples/server/server.cpp
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* * change to set
---------
Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
* ShareGPT4 compatibility (vision encoder only loading)
Load only a CLIP vision encoder (as supplied by ShareGPT finetunes)
Corrects the argument parsing for --img_mean and --img_std (which were previously not parsed but attempted to access)
Defines defaults for img_mean and img_std which are equal to the llava 1.5 CLIP encoder, so you do not have to provide them
* Update convert-image-encoder-to-gguf.py
* fix oai proxy
fix generation not stoped while bot stop talking in chat mode
fix possible `slot_id` not exist
response for cors (and pre flight)
* oai proxy: workaround for some client (such as Chatbox)
* use stop as separator to replace hardcoded `\n`
* copy to llama.cpp as subdir
* attempt enabling metal, fails
* ggml metal compiles!
* Update README.md
* initial conversion to new format, utf8 errors?
* bug fixes, but now has an invalid memory access :(
* added O3, now has insufficient memory access
* begin sync with master
* update to match latest code, new errors
* fixed it!
* fix for loop conditionals, increase result size
* fix current workflow errors
* attempt a llama.swiftui workflow
* Update .github/workflows/build.yml
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Add openai-compatible POST /v1/chat/completions API endpoint to server example
* fix code style
* Update server README.md
* Improve server README.md
* Fix server.cpp code style according to review
* server : some style changes
* server : indentation
* server : enable special tokens during tokenization by default
* server : minor code style
* server : change random string generator
* straightforward /v1/models endpoint
---------
Co-authored-by: kir-gadjello <111190790+kir-gadjello@users.noreply.github.com>
Co-authored-by: Tobi Lütke <tobi@Tobis-MacBook-Pro.local>
* llama : keep track of used KV cells + better KV cache management
* llama : zero KV cache used upon clear
ggml-ci
* llama : allow exporting a view of the KV cache (#4180)
* Allow exporting a view of the KV cache
* Allow dumping the sequences per cell in common
* Track max contiguous cells value and position as well
* Fix max contiguous empty cells index calculation
Make dump functions deal with lengths or sequences counts > 10 better
* Fix off by one error in dump_kv_cache_view
* Add doc comments for KV cache view functions
Eliminate cell sequence struct; use llama_seq_id directly
Minor cleanups
* common : add -dkvc arg for enabling kv cache dumps
---------
Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
* Support special tokens and not adding BOS to prompt in speculative
* Adapt to new should_add_bos function
* Ensure tgt and dft have same add_bos setting
