Commits

Commits on Dec 26, 2023

small typo (#28229 )
stas00
committedDec 26, 2023

Commits on Dec 12, 2023

[doc] fix typo (#27981 )
stas00
committedDec 12, 2023

Commits on Aug 28, 2023

[idefics] fix vision's hidden_act (#25787 )
stas00
committedAug 28, 2023

Commits on Aug 25, 2023

[idefics] small fixes (#25764 )
stas00
committedAug 25, 2023

Commits on Aug 24, 2023

[idefics] idefics-9b test use 4bit quant (#25734 )
stas00
committedAug 24, 2023

Commits on Aug 18, 2023

new model: IDEFICS via HuggingFaceM4 (#24796 )

committedAug 18, 2023

Commits on Jul 31, 2023

[quantization.md] fix (#25190 )
stas00
committedJul 31, 2023

Commits on Jul 24, 2023

[check_config_docstrings.py] improve diagnostics (#25012 )
stas00
committedJul 24, 2023

Commits on Jul 19, 2023

[doc] image_processing_vilt.py wrong default documented (#24931 )
stas00
committedJul 19, 2023

Commits on Jun 30, 2023

[several models] improve readability (#24585 )
stas00
committedJun 30, 2023

Commits on Jun 10, 2023

[tests] fix bitsandbytes import issue (#24151 )
stas00
committedJun 10, 2023

Commits on May 16, 2023

[AutoModel] fix torch_dtype=auto in from_pretrained (#23379 )

stas00
and
sgugger
committedMay 16, 2023

Commits on Apr 13, 2023

[trainer] update url (#22747 )
stas00
committedApr 13, 2023

Commits on Apr 12, 2023

[tests] switch to torchrun (#22712 )
stas00
committedApr 12, 2023

Commits on Mar 24, 2023

[safetensors] don't use in torch<1.10 (#22370 )
stas00
committedMar 24, 2023
[Trainer] add disclaimer that full_determinism is slow (#22368 )
stas00
committedMar 24, 2023

Commits on Mar 23, 2023

[gptj] support older pytorch version (#22325 )

committedMar 23, 2023

Commits on Mar 22, 2023

[deepspeed zero3] need generate(synced_gpus=True, ...) (#22242 )

stas00
and
sgugger
committedMar 22, 2023
[deepspeed] offload + non-cpuadam optimizer exception doc (#22044 )
stas00
committedMar 22, 2023

Commits on Mar 17, 2023

[trainer] param count for deepspeed zero3 (#22193 )
stas00
committedMar 17, 2023

Commits on Mar 14, 2023

[trainer] add --optim adamw_torch_fused for pt-2.0+ (#22144 )
stas00
committedMar 14, 2023

Commits on Mar 13, 2023

[deepspeed docs] Activation Checkpointing (#22099 )

stas00
and
sgugger
committedMar 13, 2023
[trainer] fix bug in grad accum with multiple epochs (#22098 )
stas00
committedMar 13, 2023

Commits on Mar 9, 2023

[deepspeed] offload + non-cpuadam optimizer exception (#22043 )
stas00
committedMar 9, 2023

Commits on Mar 1, 2023

[doc] deepspeed tests (#21859 )
stas00
committedMar 1, 2023

Commits on Feb 27, 2023

introduce logger.warning_once and use it for grad checkpointing code (#21804 )
stas00
committedFeb 27, 2023

Commits on Feb 23, 2023

[deepspeed tests] fix issues introduced by #21700 (#21769 )
stas00
committedFeb 23, 2023

Commits on Feb 16, 2023

[bloom] gradient_checkpointing fix (#21655 )
stas00
committedFeb 16, 2023

Commits on Feb 13, 2023

[deepspeed] performance docs (#21573 )

stas00
and
sgugger
committedFeb 13, 2023
Update setup.py (#21584 )
stas00
committedFeb 13, 2023

Commits on Feb 10, 2023

[deepspeed] deal with models w/o config.hidden_size (#21504 )
stas00
committedFeb 10, 2023
[from_pretrained] extend torch_dtype="auto" to look up config.torch_dtype first, expand docs (#21524 )
stas00
committedFeb 10, 2023

Commits on Feb 8, 2023

[tests] add missing report_to none (#21505 )
stas00
committedFeb 8, 2023

Commits on Feb 7, 2023

[tokenizer] sanitize saved config (#21483 )
stas00
committedFeb 7, 2023

Commits on Feb 6, 2023

[examples] improve block_size warning message (#21463 )
stas00
committedFeb 6, 2023