| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
|
|
|
|
|
|
|
|
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
|
|
* llama : support StableLM 2 1.6B
* convert : fix Qwen's set_vocab wrongly naming all special tokens [PAD{id}]
* convert : refactor Qwen's set_vocab to use it for StableLM 2 too
* nix : add tiktoken to llama-python-extra
* convert : use presence of tokenizer.json to determine StableLM tokenizer loader
It's a less arbitrary heuristic than the vocab size.
|
|
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>
|
|
|
|
|
|
|
|
* add safetensors support to convert-lora-to-ggml.py
* Update convert-lora-to-ggml.py
Remove white space in line 69.
|
|
Co-authored-by: Jared Van Bortel <jared@nomic.ai>
|
|
* 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>
|
|
* imatrix: speedup by avoiding unnecessary allocations and copies
* imatrix: add --no-ppl option to skip PPL calculations altogether
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
|
|
Flake lock file updates:
• Updated input 'nixpkgs':
'github:NixOS/nixpkgs/9b19f5e77dd906cb52dade0b7bd280339d2a1f3d' (2024-01-13)
→ 'github:NixOS/nixpkgs/bbe7d8f876fbbe7c959c90ba2ae2852220573261' (2024-01-19)
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
|
|
|
|
* perplexity : fix MSVC build after #5020
* try a differerent fix
|
|
ggml-ci
|
|
* Added support ccache for speedup recompilation
* cmake : option to disable ccache
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
|
|
|
|
* cuda: fix compile error in jetson platform
* cuda: update comment in ggml-cuda.cu
* cuda: update ggml-cuda.cu comment
|
|
* Fix issue with alloc causing max_compute_size to be calculated
* remove ggml_allocr_free as suggested in issue #4791
|
|
|
|
|
|
|
|
This is a relatively minor performance tweak resulting in
~10% speedup on my system.
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
|
|
* llama: add codeshell support
* llama.cpp: fix codeshell with NeoX rope
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
|
|
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
|
|
* perplexity : faster Winogrande via batching
ggml-ci
* perplexity : remove unused function
* perplexity : only tokenize selected tasks for Winogrande
|
|
* falcon arch fix for tied output embeddings
* Update llama.cpp
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update llama.cpp
* Update llama.cpp
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update llama.cpp
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
|
|
|
|
* server: defer task when no slot is available
* remove unnecessary log
---------
Co-authored-by: Xuan Son Nguyen <xuanson.nguyen@snowpack.eu>
|
|
|
|
|
|
|
|
|
|
PR #4818 (merged last week) reintroduced a config check for vocab_size that was addressed in PR #4258 (merged 2023-11-30).
Without the fix, llama2 models can't be converted. The error is:
`ValueError: The model's vocab size is set to -1 in params.json. Please update it manually. Maybe 32000?`
|
|
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>
|
|
* perplexity : faster HellaSwag
ggml-ci
* perplexity : clean-up
ggml-ci
* perplexity : no need for decode_helper
ggml-ci
* perplexity : add comments
* perplexity : option to specify max batched tasks via `n_parallel`
* perplexity : remove HellaSwag restruction for n_batch
|
|
* 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>
|
|
|
|
* Metal memory: Small memory leak on init, dangling pointer, and unused autorelease pool in graph compute
* SPM header potential fix
* Reverting symlinks
|
|
|
|
* ggml : add IQ2 to test-backend-ops + refactoring
ggml-ci
* cuda : update supports_op for IQ2
ggml-ci
* ci : enable LLAMA_CUBLAS=1 for CUDA nodes
ggml-ci
* cuda : fix out-of-bounds-access in `mul_mat_vec_q`
ggml-ci
* tests : avoid creating RNGs for each Q tensor
ggml-ci
* tests : avoid creating RNGs for each tensor
ggml-ci
|
|
* 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
|
|
* 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
* simple : no need for ggml_is_contiguous + fix bool parse
* llama : fix callback placement in llama_context_params
* backend : avoid double-ask callback calls
* simple : restore examples, imatrix will serve as a demo
|
|
* metal : create autorelease pool during library build
ggml-ci
* test : simplify
ggml-ci
|
