GitLab

Add link to pytorch CrossEntropyLoss so that one understand why '-100' is ignore by the loss function.

* fix

* fix

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* deepspeed z2/z1 state_dict bloating fix

* update

* version check

* fix str device issue

* fixup

* adapt from suggestions

* forward contrib credits from suggestions

* better fix

* added backward compatibility for older PT versions

* final fixes

* oops

* Attempting something with less branching.

---------
Co-authored-by: amyeroberts <amyeroberts@users.noreply.github.com>
Co-authored-by: Nicolas Patry <patry.nicolas@protonmail.com>

Update type annotation

fix for tpu metrics debugs string

add missing alignment heads

* Squash 88 commits

* Use markdown

* Remove mdx files due to bad rebase

* Fix modeling files due to bad rebase

* Fix style

* Update comment

* fix

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* An end to accursed version-specific imports

* No more K.is_keras_tensor() either

* Update dependency tables

* Use a cleaner call context function getter

* Add a cap to <2.14

* Add cap to examples requirements too

fix
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* Allow dict input for audio classification pipeline

* make style

* Empty commit to trigger CI

* Empty commit to trigger CI

* check for torchaudio

* add pip instructions
Co-authored-by: Sylvain <sylvain.gugger@gmail.com>

* Update src/transformers/pipelines/audio_classification.py
Co-authored-by: Nicolas Patry <patry.nicolas@protonmail.com>

* asr -> audio class

* asr -> audio class

---------
Co-authored-by: Sylvain <sylvain.gugger@gmail.com>
Co-authored-by: Nicolas Patry <patry.nicolas@protonmail.com>

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* Update src/transformers/models/whisper/modeling_tf_whisper.py
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* Update src/transformers/models/whisper/modeling_tf_whisper.py
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* fix

* fix

* fix

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* Replace python random with torch.rand to enable dynamo.export

* revert changes to flax model code

* Remove unused random import

* Fix torch template

* Move torch.manual_seed(0) to right location

* fix the grad_acc issue at epoch boundaries
Co-Authored-By: Zach Mueller <7831895+muellerzr@users.noreply.github.com>

* add contributors.

Co-authored-by: sumpster

* address comments

---------
Co-authored-by: Zach Mueller <7831895+muellerzr@users.noreply.github.com>

* fix `.to` call on 4bit models

* better check

* [AutoModel] Add AutoModelForTextEncoding

* add mt5

* add other models

* add to docs

* fix tf imports

* add tf to docs / init

* up

* fix inits

* add to dummy objects

Explain the reason to clone tensor

* fix

* fix

* Upgrade complete!

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* Refactor hyperparameter search backends

* Simpler refactoring without abstract base class

* black

* review comments:
specify name in class
use methods instead of callable class attributes
name constant better

* review comments: safer bool checking, log multiple available backends

* test ALL_HYPERPARAMETER_SEARCH_BACKENDS vs HPSearchBackend in unit test, not module. format with black.

* copyright

Fix segformer so compilation can figure out the channel dim

Update outdated hyperlink hpo_train.md 

Link to RayTune search space API docs was outdated - have provided correct new link for docs.
Co-authored-by: Joshua Samuel <66880119+Joshsamuel101@users.noreply.github.com>

fix
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

Revert "Fix gradient checkpointing + fp16 autocast for most models (#24247)"

This reverts commit 285a4801.

* fix bnb issue

* fixup

* revert and do simple patching instead

* add more details

fix
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* Slight comment cleanup

* Reduce peak mem usage when loading TF-format safetensor weights

* Tweak the PyTorch loading code to support lazy loading from safetensors

* Pass safe_open objects to the PyTorch loading function

* Do GPU transposes for speed

* One more tweak to reduce peak usage further

* One-line hasattr

* Fix bug when there's a shape mismatch

* Rename state_dict in the loading code to be clearer

* Use TF format everywhere for consistency

* [ASR pipeline] Check for torchaudio

* add pip instructions
Co-authored-by: Sylvain Gugger <sylvain.gugger@gmail.com>

---------
Co-authored-by: Sylvain Gugger <sylvain.gugger@gmail.com>

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

* fix

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

Remove redundant code

* let's go!

* initial implementation of token-level timestamps

* only return a single timestamp per token

* remove token probabilities

* fix return type

* fix doc comment

* strip special tokens

* rename

* revert to not stripping special tokens

* only support models that have alignment_heads

* add integration test

* consistently name it token-level timestamps

* small DTW tweak

* initial support for ASR pipeline

* fix pipeline doc comments

* resolve token timestamps in pipeline with chunking

* change warning when no final timestamp is found

* return word-level timestamps

* fixup

* fix bug that skipped final word in each chunk

* fix failing unit tests

* merge punctuations into the words

* also return word tokens

* also return token indices

* add (failing) unit test for combine_tokens_into_words

* make combine_tokens_into_words private

* restore OpenAI's punctuation rules

* add pipeline tests

* make requested changes

* PR review changes

* fix failing pipeline test

* small stuff from PR

* only return words and their timestamps, not segments

* move alignment_heads into generation config

* forgot to set alignment_heads in pipeline tests

* tiny comment fix

* grr

* fix

* fix

* fix

* fix

* fix

* fix

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* fix gc bug

* continue PoC on OPT

* fixes

* :exploding_head:

* fix tests

* remove pytest.mark

* fixup

* forward contrib credits from discussions

* forward contrib credits from discussions

* reverting changes on untouched files.

---------
Co-authored-by: zhaoqf123 <zhaoqf123@users.noreply.github.com>
Co-authored-by: 7eu7d7 <7eu7d7@users.noreply.github.com>

* update optimizer step for tpu

* add comment

* fix type annotation for debug arg

* fix TypeErorr

fix
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>