GitLab

Generation: only check for eos_token if set

The check for unfinished_sequences.max(), which is to find sequences
that have ended early via eos_token_id, creates a synchronization point
even when there is no eos_token, which slows inference down.

This change moves the calculation to inside the condition checking for
eos_token, so that such slowdown may be removed by disabling this token.
Co-authored-by: John Doe <john.doe@example.com>

* fix: Correct small typo in docstring

* fix: Run make fixup

fix sam example

* Patching clip model to create mask tensor on the device

* Addressing PR's comments

* Addressing PR's comments

* Addressing PR's comments

---------
Co-authored-by: Shanmugam Ramasamy <shanmugamr@shanmugamr-mlt.client.nvidia.com>

put correct link

* initial commit

* keys match

* update, fix conversion

* fixes, inference working

* fix

* more fixes

* more fixes

* clean up

* more clean up

* fix copies and add convext copied layer norm

* stash

* pretty big upfate

* cleaning

* more cleaning

* fixup stuffs

* fix copies

* fix iinit

* update test removing tokenizer

* nits

* add pretrained

* more nits

* remove tracking of pipeline

* few fixes

* update san and conversion script

* fix mask decoder and prompt encoder conversion

* fixes

* small update

* fix order

* fix

* fix image embeddings

* nites

* few fixes

* fix logits

* clean up

* fixes boxes inference

* v1 AMG

* clean up

* some clean up

* multi points support

* amg working

* fixup

* clean up

* readme

* update toctree

* fix type hint

* multiple fixes

* fixup

* fixes

* updates

* updates

* more tests

* few fixes

* change to `SamForMaskGeneration`

* doc

* fixup

* fix more tests

* multiple fixes

* fix CI tests

* refactor processor

* renamings

* draft the pipeline

* refactor

* fix tests

* fix test

* few cleanings

* fix test

* edit pipelien support chunking

* udate

* add slow tests

* fix nit

* fixup

* fix nit

* current chunk pipleine

* cast boxes in fp32

* nit

* current updates

* piepleine works

* fixup

* clean up config

* fix slow tests

* fix slow tests

* clean up

* update doc and pipeline

* adds more slow tests

* fix slow tests

* cleaning

* tests pass

* add docstring

* fix copies

* clean up

* support batch of images

* style

* dummy is needed, add tests

* fix slow tests

* fix CI

* update

* adds more tests

* fixes

* fixes

* fixup

* fixes

* few fixes

* filter

* few fixes

* some refactor

* touches finales

* fix

* style

* remove pipeline files

* fixes nits

* revert pipeline changes

* fix test

* fixup

* remove automodel for automatic mask generation

* fix failing torch tests

* update mdx

* revert removal of `MODEL_FOR_AUTOMATIC_MASK_GENERATION_MAPPING`

* update sam config based on review
Co-authored-by: amyeroberts <aeroberts4444@gmail.com>
Co-authored-by: sgugger <sylvain.gugger@gmail.com>

* update low_resolution_masks -> pred_masks
inti ln with layer_norm_eps
add_decomposed_rel_pos doc
forward doc of SamForMaskGeneration

* update processor docstring

* remove image processor import empty

* update for testing

* output vision hidden states + clean recomm
also test all iou values

* fixup

* fixup

* remove unused

* Update src/transformers/models/sam/modeling_sam.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Update src/transformers/models/sam/image_processing_sam.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* nits

* fix

* fix CI tests and slow tests

* replace with Amy's processor

* clearer docstring

* add `SamVisionNeck`

* refactor - all CI tests should pass

* fix broken import on Gcolab

* few fixes here and there

* fix another bug

* fix more bugs

* update and merge

* correct ckpt

* address comments

* add tips

* revert

* fix docstring

* replace with `SamModel`

* make fixup

* add support for bathed images and batch ed points

* make fixup this time, really

* make fixup again and again

* few fixes here and there, this should be the touche finale

* Update docs/source/en/model_doc/sam.mdx

* fixup

* correct checkpoints

* correct name

* rm unneeded file

* add notebook

---------
Co-authored-by: younesbelkada <younesbelkada@gmail.com>
Co-authored-by: amyeroberts <aeroberts4444@gmail.com>
Co-authored-by: sgugger <sylvain.gugger@gmail.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Younes Belkada <49240599+younesbelkada@users.noreply.github.com>

Fix to make sure the EOS token doesn't come back

Fixup multigpu tests

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

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

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

moved logits for m2m_100

* move preprocess_logits_for_metrics before _nested_gather in trainer.evaluation_loop

* fix

* Update src/transformers/trainer.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* fix

* fix

---------
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

fix doc comments

* Add warning about accelerate

* Version block Accelerate

* Include parse

* Apply suggestions from code review
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Check partial state

* Update param

---------
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* initial work

* Add other classes

* Refactor code

* Move warning and fix dynamic pipeline

* Issue warning when necessary

* Add test

* Do not skip auto tests

* Fix failing tests

* Refactor and address review comments

* Address review comments

* working mvp

* remove breakpoint

* fix commit

* standardize outputs

* tmp commit

* tests almost ready

* tmp commit

* skip a few models

* Add streaming; Docs and examples

* document limitations

* PR commits

* Amy PR comments

* fix

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

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

docs: ko: fix anchor links for docs (auto_tutorial, training)
Co-authored-by: Hyeonseo Yun <0525_hhgus@naver.com>
Co-authored-by: Sohyun Sim <96299403+sim-so@users.noreply.github.com>
Co-authored-by: Na Yeon Han <nayeon2.han@gmail.com>
Co-authored-by: Wonhyeong Seo <wonhseo@kakao.com>
Co-authored-by: Jungnerd <46880056+jungnerd@users.noreply.github.com>

* wrong argument name

* append eos_token_id

* all tokenizers need mask and ctc_blank tokens

* remove reduction factor from feature extractor

* add proper TTS loss

* did shifting the wrong way around

* mask out padded portions

* remove logits again (don't really need it)

* fix unit tests

* fixup

* pad also returns the decoder attention mask, since that's useful to have

* clean up feature extractor logic

* pad can handle TTS task too

* remove stop_labels from loss calculation

* simplify logic

* fixup

* do -100 masking properly

* small STFT optimization (calculate mel filterbanks only once)

* replace torchaudio fbanks with audio_utils

* remove torchaudio dependency

* simplify & speed up the STFT

* don't serialize window and mel filters

* output cross attentions when generating speech

* add guided attention loss

* fix failing test

* Update src/transformers/models/speecht5/feature_extraction_speecht5.py
Co-authored-by: Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>

* Update src/transformers/models/speecht5/modeling_speecht5.py
Co-authored-by: Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>

* change type annotation of attention_mask to LongTensor

* extract loss into class

* remove unused frame_signal_scale argument

* use config object in loss class

* fix type annotations in doc comments

* change optional to just bool

* implement missing tokenizer method

* add deprecation warning

* Update src/transformers/models/speecht5/feature_extraction_speecht5.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Update src/transformers/models/speecht5/feature_extraction_speecht5.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* add deprecation warning for stop_labels

---------
Co-authored-by: Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Mark auto models as important

* Annoying file with bad line endings

* Use accelerate for device management

* Add accelerate to setup
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

Revert "Use code on the Hub from another repo (#22698)"

This reverts commit ea7b0a53.

* Simplify update metadata job

* Match more branch names

* Install all what is necessary

* Install all what is necessary

* Forgot the dev

* Install less stuff

* This syntax?

Remove accelerate from tf

fix-squeeze-tuple

* fix

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

* initial work

* Add other classes

* Refactor code

* Move warning and fix dynamic pipeline

* Issue warning when necessary

* Add test

docs: ko: tasks/translation.mdx

fix: docs: ko: sagemaker anchors and  `_toctree.yml`
Co-authored-by: Hyeonseo Yun <0525_hhgus@naver.com>
Co-authored-by: Gabriel Yang <gabrielwithhappy@gmail.com>
Co-authored-by: Sohyun Sim <96299403+sim-so@users.noreply.github.com>
Co-authored-by: Na Yeon Han <nayeon2.han@gmail.com>
Co-authored-by: Wonhyeong Seo <wonhseo@kakao.com>

docs: ko: translated `custom_models.mdx`
Co-authored-by: Wonhyeong Seo <wonhseo@kakao.com>
Co-authored-by: Gabriel Yang <gabrielwithhappy@gmail.com>
Co-authored-by: Jungnerd <46880056+jungnerd@users.noreply.github.com>

* fix

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