huggingface/transformers

Pull Requests Commits

Fix: Disable torch.autocast in RotaryEmbedding of Gemma and LLaMa for MPS device (#29439)

currybab committed 1 year ago

Verified d45f47ab

Substantially reduce memory usage in _update_causal_mask for large batches by using .expand instead of .repeat [needs tests+sanity check] (#29413)

nqgl committed 1 year ago

Verified 2a939f20

Fix `TextGenerationPipeline.__call__` docstring (#29491)

alvarobartt committed 1 year ago

Verified 965cf677

added the max_matching_ngram_size to GenerationConfig (#29131)

mosheber committed 1 year ago

Verified 19fb1e22

Generate: torch.compile-ready generation config preparation (#29443)

gante committed 1 year ago

Verified ddb4fda3

Fix test failure on DeepSpeed (#29444)

muellerzr committed 1 year ago

Verified 9322576e

Avoid dummy token in PLD to optimize performance (#29445)

ofirzaf committed 1 year ago

Verified 0a5b0516

Generate: get generation mode from the generation config instance 🧼 (#29441)

gante committed 1 year ago

Verified 700d48fb

Generate: add tests for caches with `pad_to_multiple_of` (#29462)

gante committed 1 year ago

Verified 41f7b7ae

Fix TrainingArguments regression with torch <2.0.0 for dataloader_prefetch_factor (#29447)

Matthew Hoffman committed 1 year ago

Verified 2890116a

[`docs`] Add starcoder2 docs (#29454)

younesbelkada committed 1 year ago

Verified b27aa206

[`Docs` / `Awq`] Add docs on exllamav2 + AWQ (#29474)

younesbelkada committed 1 year ago

Verified 2a002d07

[FIX] `offload_weight()` takes from 3 to 4 positional arguments but 5 were given (#29457)

faaany committed 1 year ago

Verified 00bf4427

🌐 [i18n-KO] Translated generation_strategies.md to Korean (#29086)

AI4Harmony committed 1 year ago

Verified 7b01579f

[i18n-zh] Translate add_new_pipeline.md into Chinese (#29432)

windsonsea committed 1 year ago

Verified 638c423c

Automatic safetensors conversion when lacking these files (#29390)

LysandreJik committed 1 year ago

Verified a69cbf4e

Update pytest `import_path` location (#29154)

loadams committed 1 year ago

Verified 9c5e5609

Fix bug with passing capture_* args to neptune callback (#29041)

AleksanderWWW committed 1 year ago

Verified 8f3f8e67

[`Add Mamba`] Adds support for the `Mamba` models (#28094)

ArthurZucker committed 1 year ago

Verified fb1c62e9

Generate: inner decoding methods are no longer public (#29437)

gante committed 1 year ago

Verified 87a0783d

[`Udop imports`] Processor tests were not run. (#29456)

ArthurZucker committed 1 year ago

Verified 4d892b72

Revert-commit 0d52f9f582efb82a12e8d9162b43a01b1aa0200f (#29455)

ArthurZucker committed 1 year ago

Verified 57d007b9

ArthurZucker committed 1 year ago

0d52f9f5

[`UdopTokenizer`] Fix post merge imports (#29451)

ArthurZucker committed 1 year ago

Verified 13285220

[tests] enable test_pipeline_accelerate_top_p on XPU (#29309)

faaany committed 1 year ago

Verified fa7f3cf3

[docs] Update starcoder2 paper link (#29418)

xenova committed 1 year ago

Verified ebccb091

Fix max length for BLIP generation (#29296)

zucchini-nlp committed 1 year ago

Verified bd891aed

Exllama kernels support for AWQ models (#28634)

IlyasMoutawwakil committed 1 year ago

Verified 4fc708f9

FIX [`Generation`] Fix some issues when running the MaxLength criteria on CPU (#29317)

younesbelkada committed 1 year ago

Verified 81c8191b

[Docs] Spanish Translation -Torchscript md & Trainer md (#29310)

njackman-2344 committed 1 year ago

Verified e9476832

Newer Older