huggingface/transformers

Pull Requests Commits

ArthurZucker committed 1 year ago

092f1fda

fix merge conflicts between llama and gemma

ArthurZucker committed 1 year ago

bf5163f4

Use `torch.bool` instead of `torch.int64` for non-persistant causal mask buffer (#29241)

fxmarty committed 1 year ago

6c45f0f6

Patch YOLOS and others (#29353)

NielsRogge committed 1 year ago

bfefb8ef

RoPE loses precision for Llama / Gemma + Gemma logits.float() (#29285)

danielhanchen committed 1 year ago

20164cc2

[`Llama ROPE`] Fix torch export but also slow downs in forward (#29198)

ArthurZucker committed 1 year ago

d5ec1946

[`T5 and Llama Tokenizer`] remove warning (#29346)

ArthurZucker committed 1 year ago

bf99e862

Improve _update_causal_mask performance (#29210)

alessandropalla committed 1 year ago

6d023503

FIX [`Gemma`] Fix bad rebase with transformers main (#29170)

younesbelkada committed 1 year ago

4f8689e8

Release: v4.38.1

ArthurZucker committed 1 year ago

a0857740

[Gemma] Fix eager attention (#29187)

sanchit-gandhi committed 1 year ago

2f54e0b3

[ `gemma`] Adds support for Gemma 💎 (#29167)

ArthurZucker committed 1 year ago

08ab54ad

[`Maskformer`] safely get backbone config (#29166)

amyeroberts committed 1 year ago

2de93141

🚨 Llama: update rope scaling to match static cache changes (#29143)

gante committed 1 year ago

476957b5

Release: 4.38.0

ArthurZucker committed 1 year ago

7a4bec6e

Add support for fine-tuning CLIP-like models using contrastive-image-text example (#29070)

tjs-intel committed 1 year ago

Verified ee3af60b

Revert low cpu mem tie weights (#29135)

amyeroberts committed 1 year ago

Verified 0996a100

[`Core tokenization`] `add_dummy_prefix_space` option to help with latest issues (#28010)

ArthurZucker committed 1 year ago

Verified 15cfe389

FIX [`PEFT` / `Trainer` ] Handle better peft + quantized compiled models (#29055)

younesbelkada committed 1 year ago

Verified efdd4366

[`cuda kernels`] only compile them when initializing (#29133)

ArthurZucker committed 1 year ago

Verified 5e95dcab

Generate: unset GenerationConfig parameters do not raise warning (#29119)

gante committed 1 year ago

Verified a7755d24

Llama: fix batched generation (#29109)

gante committed 1 year ago

Verified 7d312ad2

FIX [`bnb` / `tests`] Propagate the changes from #29092 to 4-bit tests (#29122)

younesbelkada committed 1 year ago

Verified ff76e7c2

Abstract image processor arg checks. (#28843)

molbap committed 1 year ago

Verified 1c9134f0

FEAT [`Trainer` / `bnb`]: Add RMSProp from `bitsandbytes` to HF `Trainer` (#29082)

younesbelkada committed 1 year ago

Verified f7ef7cec

Move misplaced line (#29117)

kno10 committed 1 year ago

Verified a7ff2f23

[`gradient_checkpointing`] default to use it for torch 2.3 (#28538)

ArthurZucker committed 1 year ago

Verified 9094abe8

Fixed nll with label_smoothing to just nll (#28708)

nileshkokane01 committed 1 year ago

Verified 49c0b293

storing & logging gradient norm in trainer (#27326)

shijie-wu committed 1 year ago

Verified 4f09d0fd

Fix two tiny typos in `pipelines/base.py::Pipeline::_sanitize_parameters()`'s docstring (#29102)

sadra-barikbin committed 1 year ago

Verified a4851d94

Older