GitLab

* fix: half inference error

norm_factor is still torch.float32 after using model.half

So I changed it to register_buffer so I can change it to torch.float16 after using model.half

* fix: Added a variable "persistent=False"

* run make style

* expose

* style

* add dummy object

* amazed by the quality of transformers CI

* fix code not exportable

* fix

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

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

* Seq2SeqTrainingArgs overriding to_dict for GenerationConfig json support

* seq2seqTrainingArgs to_dict calling super method before handling genconf

fix bug of CLAP: https://github.com/LAION-AI/CLAP/issues/62

Updates Swin MIM output class to match other masked image modeling outputs

* fix

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

* support gpt neox generate with inputs embeds

* Update src/transformers/models/gpt_neox/modeling_gpt_neox.py

great thx for the suggestion!
Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

---------
Co-authored-by: Lei Li <tobiaslee@qq.com>
Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

* integration tests were not being run

* add tests for short input waveform

* rewrite test for long input

* even more betterer

* my bad

* oh boy

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>

This predicate will become available in tensorflow starting with version
2.14.
Co-authored-by: Russell Power <power@google.com>

* cleanup

* updates

* more refactoring

* make style

* update inits

* support other inputs in base

* update based on review
Co-authored-by: Nicolas Patry <patry.nicolas@gmail.com>

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

* update

* fixup

* TODO x and y to refactor, _h _w refactored here

* update docstring

* more nits

* style on these

* more doc fix

* rename variables

* update

* updates

* style

* update

* fix `_mask_to_rle_pytorch`

* styling

* fix ask to rle, wrong outputs

* add device arg

* update

* more updates, fix tets

* udpate

* update docstrings

* styling

* fixup

* add notebook on the docs

* update orginal sizes

* fix docstring

* updat condition on point_per-batch

* updates tests

* fix CI  test

* extend is required, append does not work!

* fixup

* fix CI tests

* whit pixels left

* address doc comments

* fix doc

* slow pipeline tests

* update auto init

* add revision

* make fixup

* update p!ipoeline tag when calling tests

* alphabeitcal order in inits

* fix copies

* last style nits

* Apply suggestions from code review
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* reformat docstring

* more reformat

* address most of the comments

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

* final refactor

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

* fixup and fix slow tests

* revert

---------
Co-authored-by: Nicolas Patry <patry.nicolas@gmail.com>
Co-authored-by: Nicolas Patry <patry.nicolas@protonmail.com>
Co-authored-by: younesbelkada <younesbelkada@gmail.com>
Co-authored-by: Younes Belkada <49240599+younesbelkada@users.noreply.github.com>
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* Pin optax version

* Pin flax too

* Fixup

Fix weight tying in ESM

* moved labels to the same device as logits for OTP model

* moved labels to the same device as logits for CODEGEN model

* Update modeling_codegen.py

* moved labels to the same device as logits for gptj and pix2struct model

* Update modeling_pix2struct.py

minor refactoring to allow compatible datasets to work.

change to `facebook/sam-vit-base`

* fix

* Update .github/workflows/self-nightly-past-ci-caller.yml
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

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

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>