GitLab

* move scaling to nn.Module

* let the test be here for now (need to fix)

* failing tests

* last failing models

* Revert commit 4c14817f38

* clean-up

* oops forgot

* codestyle

* raise NotImplemented when possible

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

* skip tests in respective modeling files

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

* [test_all] Remove static pretrained maps from the library's internals

* Deprecate archive maps instead of removing them

* Revert init changes

* [test_all] Deprecate instead of removing

* [test_all] PVT v2 support

* [test_all] Tests should all pass

* [test_all] Style

* Address review comments

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

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

* [test_all] trigger tests

* [test_all] LLAVA

* [test_all] Bad rebase

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

* fix stablelm dropout argument type error

* fix docs of _flash_attention_forward

* fix all docs of _flash_attention_forward

* fix docs of _flash_attention_forward in starcoder2

---------
Co-authored-by: oliang <oliang@tencent.com>

* fix fa2 autocasting when using quantization

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

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

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

* add sdpa

* wip

* cleaning

* add ref

* yet more cleaning

* and more :)

* wip llama

* working llama

* add output_attentions=True support

* bigcode sdpa support

* fixes

* gpt-bigcode support, require torch>=2.1.1

* add falcon support

* fix conflicts falcon

* style

* fix attention_mask definition

* remove output_attentions from attnmaskconverter

* support whisper without removing any Copied from statement

* fix mbart default to eager renaming

* fix typo in falcon

* fix is_causal in SDPA

* check is_flash_attn_2_available in the models init as well in case the model is not initialized through from_pretrained

* add warnings when falling back on the manual implementation

* precise doc

* wip replace _flash_attn_enabled by config.attn_implementation

* fix typo

* add tests

* style

* add a copy.deepcopy on the config in from_pretrained, as we do not want to modify it inplace

* obey to config.attn_implementation if a config is passed in from_pretrained

* fix is_torch_sdpa_available when torch is not installed

* remove dead code

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

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

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

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

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

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

* remove duplicate pretraining_tp code

* add dropout in llama

* precise comment on attn_mask

* add fmt: off for _unmask_unattended docstring

* precise num_masks comment

* nuke pretraining_tp in LlamaSDPAAttention following Arthur's suggestion

* cleanup modeling_utils

* backward compatibility

* fix style as requested

* style

* improve documentation

* test pass

* style

* add _unmask_unattended tests

* skip meaningless tests for idefics

* hard_check SDPA requirements when specifically requested

* standardize the use if XXX_ATTENTION_CLASSES

* fix SDPA bug with mem-efficient backend on CUDA when using fp32

* fix test

* rely on SDPA is_causal parameter to handle the causal mask in some cases

* fix FALCON_ATTENTION_CLASSES

* remove _flash_attn_2_enabled occurences

* fix test

* add OPT to the list of supported flash models

* improve test

* properly test on different SDPA backends, on different dtypes & properly handle separately the pad tokens in the test

* remove remaining _flash_attn_2_enabled occurence

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

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

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

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

* Update docs/source/en/perf_infer_gpu_one.md
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

* remove use_attn_implementation

* fix docstring & slight bug

* make attn_implementation internal (_attn_implementation)

* typos

* fix tests

* deprecate use_flash_attention_2=True

* fix test

* add back llama that was removed by mistake

* fix tests

* remove _flash_attn_2_enabled occurences bis

* add check & test that passed attn_implementation is valid

* fix falcon torchscript export

* fix device of mask in tests

* add tip about torch.jit.trace and move bt doc below sdpa

* fix parameterized.expand order

* move tests from test_modeling_attn_mask_utils to test_modeling_utils as a relevant test class is already there

* update sdpaattention class with the new cache

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

* Update src/transformers/models/bark/modeling_bark.py

* address review comments

* WIP torch.jit.trace fix. left: test both eager & sdpa

* add test for torch.jit.trace for both eager/sdpa

* fix falcon with torch==2.0 that needs to use sdpa

* fix doc

* hopefully last fix

* fix key_value_length that has no default now in mask converter

* is it flacky?

* fix speculative decoding bug

* tests do pass

* fix following #27907

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

* support FA2

* fix typo

* fix broken tests

* fix more test errors

* left/right

* fix bug

* more test

* typo

* fix layout flash attention falcon

* do not support this case

* use allclose instead of equal

* fix various bugs with flash attention

* bump

* fix test

* fix mistral

* use skiptest instead of return that may be misleading

* add fix causal arg flash attention

* fix copies

* more explicit comment

* still use self.is_causal

* fix causal argument

* comment

* fixes

* update documentation

* add link

* wrong test

* simplify FA2 RoCm requirements

* update opt

* make flash_attn_uses_top_left_mask attribute private and precise comment

* better error handling

* fix copy & mistral

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

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

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

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

* use is_flash_attn_greater_or_equal_2_10 instead of is_flash_attn_greater_or_equal_210

* fix merge

* simplify

* inline args

---------
Co-authored-by: Felix Marty <felix@hf.co>
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* add whisper fa2

* correct

* change all

* correct

* correct

* fix more

* fix more

* fix more

* fix more

* fix more

* fix more

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

* fix more

* fix more

* fix more

* fix more

* fix more

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

* [FA2 Bart] Add FA2 to all Bart-like

* better

* Refactor attention mask

* remove all customized atteniton logic

* format

* mass rename

* replace _expand_mask

* replace _expand_mask

* mass rename

* add pt files

* mass replace & rename

* mass replace & rename

* mass replace & rename

* mass replace & rename

* Update src/transformers/models/idefics/modeling_idefics.py

* fix more

* clean more

* fix more

* make style

* fix again

* finish

* finish

* finish

* finish

* finish

* finish

* finish

* finish

* finish

* finish

* Apply suggestions from code review

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

* small fix mistral

* finish

* finish

* finish

* finish

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

* fix

* more fixes

* fix other models

* fix long t5

* use `gradient_checkpointing_func` instead

* fix copies

* set `gradient_checkpointing_func` as a private attribute and retrieve previous behaviour

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

* replace it with `is_gradient_checkpointing_set`

* remove default

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

* fixup

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

* v1

* fix

* remove `create_custom_forward`

* fixup

* fixup

* add test and fix all failing GC tests

* remove all remaining `create_custom_forward` methods

* fix idefics bug

* fixup

* replace with `__call__`

* add comment

* quality

In assisted decoding, pass model_kwargs to model's forward call (fix prepare_input_for_generation in all models) (#25242)

* In assisted decoding, pass model_kwargs to model's forward call

Previously, assisted decoding would ignore any additional kwargs
that it doesn't explicitly handle. This was inconsistent with other
generation methods, which pass the model_kwargs through
prepare_inputs_for_generation and forward the returned dict to the
model's forward call.

The prepare_inputs_for_generation method needs to be amended in all
models, as previously it only kept the last input ID when a past_key_values
was passed.

* Improve variable names in _extend_attention_mask

* Refactor extending token_type_ids into a function

* Replace deepcopy with copy to optimize performance

* Update new persimmon model with llama changes for assisted generation

* Update new mistral model for assisted generation with prepare_inputs_for_generation

* Update position_ids creation in falcon prepare_inputs_for_generation to support assisted generation

* fix_mbart_tied_weights

* add test

* Fix GPTNeoX beam search when using parallelize

* Fix beam search idx device when using model parallel

* remove onnx related stuff
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

* fix: move test_beam_search_on_multi_gpu to GenerationTesterMixin

* fix: add right item to _no_split_modules of MegaPreTrainedModel

* fix: add num_beams within parallelized beam_search test
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

---------
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* fix

* revert cahnges and update resizing of embedding layer

* use wraning

* fixup

* more styling nits

* fix all tests that overload the embedding tests

* 👀👀 remove breakpoint

* remove useless overload + overload correctly where needed

* resize lm head with new vocab size

* reverse not necessary changes

* style

* fix CIs!

* fix last CI tests, adapt bark and Marian

* fixup

* Update modeling_llama.py

Removing unnecessary `device=device`

* fix in all occurrences of _make_causal_mask

* fix annotations

* fix copies

* Preliminary work on some models

* Fix test load missing and make sure nonpersistent buffers are tested

* Always ignore nonpersistent buffers if in state_dict

* Treat models

* More models

* Treat remaining models

* Fix quality

* Fix tests

* Remove draft

* This test is not needed anymore

* Fix copies

* Fix last test

* Newly added models

* Fix last tests

* Address review comments

* fix

* fix

* fix

* fix

* fix

* fix

---------
Co-authored-by: ydshieh <ydshieh@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

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

This reverts commit 285a4801.

* 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>

* First test

* Add info for all models

* style

* Repo consistency

* Fix last model and cleanup prints

* Repo consistency

* Use consistent function for detecting tied weights

feat(model parallelism): moving the labels to the same device as the logits for gpt2 and bart (#22591)

Fix docs and doctests

* ensure causal_mask is created directly on device

* add copy tag to opt, update bart implementation

* add device to all _make_causal_mask copies

* formatting fixes

* more manual fixes due to unlinked versions of _prepare_decoder_attention_mask

add `accelerate` support for MBart

* Change the .view call to .reshape

* Change the .view call to .reshape to all the copies from bart attention

* Fix copies and style

* Fix copies and style

* Fix copies and style

* Fix copies and style

* Fix copies and style

* Revert unneccessary changes

* Revert unneccessary changes

* Revert unneccessary changes

* Revert unneccessary changes

* logger.warning_once

* style

fix typo

* Update document of WhisperDecoderLayer

* Update modeling_mbart.py

* Update doc with utils/check_copies.py --fix_and_overwrite

* Update modeling_xlm_prophetnet.py

* Enforce single model initialization

* Add OneFormer example for problem 3

* Do it the Stas way

* Actually rename the uses...

* Rewrite test

* Try to change the test this way

* Fix all init slow/fast tests

* Break connection

* Fix more tests

* Fix test for initialization

* Remove custom test

* Quality

* Fix last failing tests

* The end?

* fix past renamed to past_key_value

* update more `past`that were ski^êd

* fixup

* remove changes made to rag

* refactor `_reorder_cache` to use `past_key_values`

* fix git `prepare_inputs_for_generation` to pass tests when false is needed in use_cache

* Result of black 23.1

* Update target to Python 3.7

* Switch flake8 to ruff

* Configure isort

* Configure isort

* Apply isort with line limit

* Put the right black version

* adapt black in check copies

* Fix copies

* Clean all models

* Style

* Last to remove

* address review comments

* Address review comments

* Remove all hf-internal-testing checkpoints that can be removed

* Fix copies

* Put back processor_class in TF example

* Address review comment

* start cleanup

* more updates

* more models are affected

* more updates

* update generation utils

* style

* revert change that removed reorder cachce

* update generation utils

* style

* style

* remove reorder cache