GitLab

* init

* test

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

Co-authored-by: Joao Gante <joao@huggingface.co>

fix test

only test input_embeds, not decoder_input_embeds

* fix

* fix

* fix

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

* symbolic trace supports inputs_embeds

* fix test?

* Update tests/test_modeling_common.py
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

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

* Rebase to main

* Fix attention implementation autoset for tex and vision configs

* Fixup

* Minor fixes

* Fix copies

* Fix attention_mask for FA2

* Add eqvivalence tests for siglip

* Remove right padding test

* Uncomment flaky

* Fix import

* Add to docs

* Fix test message

* Add sdpa

* Add sdpa equivalence test

* Add siglip sdpa to docs

* Fix typing for attention output

* Add sdpa tests

* Fix signature of FA2

* Autoset attn_implementation in config

* Rename bsz -> batch_size

* Move back autoset attn method

* Mark as flaky

* Correct attention mask padding

* [run-slow] siglip

* Add FA2 and sdpa docs

* Style fix

* Remove flaky for FA2 test

* Change attention implementation set

* Change attn_implementaiton propogation

* Fix typos

* Add modality to assert message

* Add more sdpa backends in test

* [run slow] siglip

* Add math sdpa backend for all options

* [run slow] siglip

* First draft

* Add docs

* Clean up code

* Convert model

* Add image processor

* Convert Zoe_K

* More improvements

* Improve variable names and docstrings

* Improve variable names

* Improve variable names

* Replace nn.sequential

* More improvements

* Convert ZoeD_NK

* Fix most tests

* Verify pixel values

* Verify pixel values

* Add squeeze

* Update beit to support arbitrary window sizes

* Improve image processor

* Improve docstring

* Improve beit

* Improve model outputs

* Add figure

* Fix beit

* Update checkpoint

* Fix repo id

* Add _keys_to_ignore_on_load_unexpected

* More improvements

* Address comments

* Address comments

* Address comments

* Address comments

* Rename variable name

* Add backbone_hidden_size

* Vectorize

* Vectorize more

* Address comments

* Clarify docstring

* Remove backbone_hidden_size

* Fix image processor

* Remove print statements

* Remove print statement

* Add integration test

* Address comments

* Address comments

* Address comments

* Address comments

* Add requires_backends

* Clean up

* Simplify conversion script

* Simplify more

* Simplify more

* Simplify more

* Clean up

* Make sure beit is loaded correctly

* Address comment

* Address bin_configurations

* Use bin_configurations

* Convert models, add integration tests

* Fix doc test

* Address comments

* Unify regressor classes

* Clarify arguments

* Improve resize_image

* Add num_relative_features

* Address comment

* [run-slow]beit,data2vec,zoedepth

* [run-slow]beit,data2vec,zoedepth

* Address comments

* Address comment

* Address comment

* Replace nn.TransformerEncoderLayer and nn.TransformerEncoder

* Replace nn.MultiheadAttention

* Add attributes for patch transformer to config

* Add tests for ensure_multiple_of

* Update organization

* Add tests

* [run-slow] beit data2vec

* Update ruff

* [run-slow] beit data2vec

* Add comment

* Improve docstrings, add test

* Fix interpolate_pos_encoding

* Fix slow tests

* Add docstring

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

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

* Improve tests and docstrings

* Use run_common_tests

* Improve docstrings

* Improve docstrings

* Improve tests

* Improve tests

* Remove print statements

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

* fix galore lr display with lr schedulers

* style

* add some tests to check for displayed lrs

* copy-paste err for warmup steps

* standardize the default lr to be only in the optimizer

* trying out my luck with the reads

* cast image features to model.dtype where needed to support FP16 or other precision in pipelines

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

* Use .to instead

* Add FP16 pipeline support for zeroshot audio classification

* Remove unused torch imports

* Add docs on FP16 pipeline

* Remove unused import

* Add FP16 tests to pipeline mixin

* Add fp16 placeholder for mask_generation pipeline test

* Add FP16 tests for all pipelines

* Fix formatting

* Remove torch_dtype arg from is_pipeline_test_to_skip*

* Fix format

* trigger ci

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

* Add siglip loss function

* Update docs

* Enable training tests
[experimental] enable GC training tests as it has worked for my own data

* Remove test_training* overrides to enable training tests
[run_slow] siglip

* Skip training tests for Siglip text model and ImageClassificationModel
[run_slow] siglip

* Skip GC training tests for SiglipForImageClassification

* Explicitly skip training tests for SiglipVisionModel
Add skip reason for training tests for SiglipTextModel

* Remove copied from to fix CI

* Code agent: allow function persistence between steps

* skip 3 7b tests

* fix

* fix

* fix

* [run-slow] gemma

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

* Fix serialization

* style

* add test

* Fix init for rt-detr heads

* Fixup

* Add separate prior_prob value to config for initialization

* Add bbox init

* Change to 1 / num_labels init

* Adjust weights init test

* Fix style for test

* Fix cache and type conversion

* Add test

* Fixup

* nit

* [run slow] rt_detr

* Fix test

* Fixup

* [run slow] rt_detr

* Update src/transformers/models/rt_detr/modeling_rt_detr.py

gemma 2 slow tests

* fix assisted decoding

* check None

* fix typo

* fix _prepare_special_tokens

* fix style

* fix lint

* add tests for assisted decoding

* fix style

* fix tests check

Make the order of array items consistent using sorted()

* rely on the tokenizer default kwargs

* fix a few tests

* make work with cache abstraction

* correct for static cache

* hacks for compile

* make fast

* fix

* fix pos ids

* generate

* fix sdpa

* fix sdpa cache pos

* fix fa2

* clean fa2

* integrate cache into generate

* make style

* copies

* more copies

* update eager

* update sdpa

* update fa2

* simplify

* use cache pos

* always compute cross-cache for debug

* avoid recompiles
Co-authored-by: Arthur Zucker <arthur@huggingface.co>

* fix fix

* fix fix fix

* more fix

* try encoder-decoder cache (too messy)

* revert encoder-decoder cache

* check cross-attn cache

* use enc-dec dataclass

* use richer enc-dec dataclass

* clean-up

* revert static cache changes

* small fixes

* revert to cpu flag

* fix copies

* add static slow test

* past k/v docstring

* more docstrings

* cache_position docstrings

* add to docs

* add enc-dec cache to docs

* make style

* fix after rebase

* fix beam

* style

* fix generation strategies

* fix most decoder-only tests

* style

* skip test

* more clean up

* small docstrings

* Apply suggestions from code review
Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

* add todo

* only crop self-attn

* check cache in mixin

* style

* fix re-compile after rebase

* move `is_updated` logic to enc-dec wrapper

* revert back

* revert cache back

* finalise design

* fix

* fix fix

* style

* Update src/transformers/cache_utils.py
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

* deprecate

* updates

* final updates

* style

* style

---------
Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

* move

* move

* move

* move

* Update tests/utils/test_image_processing_utils.py
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>

* add gather_use_object arguments

* fix name and pass the CI test for Seq2SeqTrainer

* make style

* make it to functools

* fix typo

* add accelerate version:

* adding warning

* Update src/transformers/trainer.py
Co-authored-by: Marc Sun <57196510+SunMarc@users.noreply.github.com>

* make style

* Update src/transformers/training_args.py

* check function move to initial part

* add test for eval_use_gather_object

---------
Co-authored-by: Marc Sun <57196510+SunMarc@users.noreply.github.com>

* fix: use return_dict parameter

* fix: type checks

* fix: unused imports

* update: one-line if else

* remove: recursive check

* nit

* toctree issue

* protect gemma2 tests as well

* sdpa supported

* inital commit

* Add doc

* protect?

* fixup stuffs

* update tests

* fix build documentation

* mmmmmmm config attributes

* style

* nit

* uodate

* nit

* Fix docs

* protect some stuff

---------
Co-authored-by: Lysandre <lysandre@huggingface.co>

* change anchor_image_size None for compatibility

* make fix-copies

* Allow dtype str for torch_dtype in from_pretrained

* Update docstring

* Add tests for str torch_dtype

* Skip tests properly

* [test_all]

* Add 'reason' as kwarg for skipTest

* [test_all] Fix up

* [test_all]

* Fix dtype casting in modeling_swin2sr to allow non-FP32 inference

* Fix formattting

* Fix for swinv2 too

* Update src/transformers/models/swin2sr/modeling_swin2sr.py
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* Update src/transformers/models/swinv2/modeling_swinv2.py
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* Add FP16 tests for swin2sr and swinv2

* [run_slow] swin2sr, swinv2

* [run_slow] swin2sr, swinv2

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

* fix extended attention mask

* add slow test for detection instance

* [run-slow]paligemma

* squash into single commit

* run diff once more

* docstring

* tests

* minor chnages and ready to go

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

* Update tests/models/vipllava/test_modeling_vipllava.py
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* [run-slow] llava-next-video

* [run-slow] llava-next-video

* [run-slow] llava_next_video

* fix two tests

* fix slow tests

* remove logit checks due to numeric errors

* run test once more

* [run-slow] llava_next_video

* final try to pass the test

* [run-slow] llava_next_video

* [run-slow] llava_next_video

* [run-slow] llava_next_video

* style

* fix

* style

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

* [run_slow] rt_detr

* Fix positional embeddings and anchors dtypes

* [run slow] rt_detr

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

* Fixup

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

* fix llama fsdp

* fixup

* adding FSDP tests for CPU offloading

* fixes

* fix tests

* fix tests

* add it for mixtral

* propagate the changes on other models

* Update src/transformers/models/phi/modeling_phi.py

* Delete utils/testing_scripts/fsdp_cpu_offloading.py

Remove script - FSDP + CPU offloading it tested in the test suite

* Delete utils/testing_scripts/dummy_fsdp_config.yml

* Update + add cache_positions docstring

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

* starting support for sdpa in `gptneox` models

* small comment on tests

* fix dropout

* documentation and style

* clarify concrete paths for reference

* generalise attn projections and rope application

added head mask check to sdpa mask creation

handle sdpa memory backend bug via own version flag

* update docs and style

* move dtype casting outside of general attn_projection_and_rope function

fix flash_attn_2 stuff

* more generic attn warning if output_attns or head_mask

* simplify head mask check by moving head mask creation to a later point

* remove copied llama artifact

* remove padding_mask from attention function signature

* removing unnecessary comments, only "save" attn implementation once

* [run_slow] gpt_neox

* Add ViTImageProcessor to tests

* Correct data format

* Review comments

* squash in single commit

* add docs

* dummy obj

* more changes in diff converter

* tiny fix

* make docs happy

* skip test

* repo consistency tests

* update docstring

* style

* fix tests

* change diff imports

* [run-slow] instructblipvideo

* [run-slow] instructblipvideo

* fix tests and remove logit check

* [run-slow] instructblipvideo

* fix output data type of image classification

* add tests for low-precision pipeline

* add bf16 pipeline tests

* fix bf16 tests

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

* fix import

* fix import torch

* fix style

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

* device-map siglip

* move split modules to PretrainedSigLip

* fix bug about add_special_tokens and so on

* improve add_special_tokens and padding behavior

* add a test case for add_special_tokens and padding