* Updated docstring for _determine_best_metric.
* Updated docstring for metric_for_best_model.
* Added test case for save strategy.
* Updated incorrect test case.
* Changed eval_strategy to match save_strategy.
* Separated test cases for metric.
* Allow load_best_model when save_strategy == "best".
* Updated docstring for metric_for_best_model.
* fix: processing odd number of frames
* feat: add test case
* update: test one frame
* feat: support custom patch size
* fix: test with videos
* revert: change on patch repeat
* fix: much wow
* update: fixups
* fixup pls
* ruff fixup
* fix typo at least
* Correctly list the chat template file in the saved files list
* Update src/transformers/tokenization_utils_base.py
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
* Add save file checking to test
* make fixup
* better filename handling
* make fixup
---------
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
* add audio_token attribute to proc
* expand input_ids
* and legacy and expanded input_ids
* test update
* split lines
* add possibility not to provide eos and bos audio tokens
* raise errors
* test incorrect number of audio tokens
* add example
* fmt
* typo
* first adding diffllama
* add Diff Attention and other but still with errors
* complate make attention Diff-Attention
* fix some bugs which may be caused by transformer-cli while adding model
* fix a bug caused by forgetting KV cache...
* Update src/transformers/models/diffllama/modeling_diffllama.py
You don't need to divide by 2 if we use same number of attention heads as llama. instead you can just split in forward.
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* Update src/transformers/models/diffllama/modeling_diffllama.py
fit to changeing "num_heads // 2" place
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* Update src/transformers/models/diffllama/modeling_diffllama.py
new codes are more meaningful than before
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* Update src/transformers/models/diffllama/modeling_diffllama.py
new codes are more meaningful than before
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* Update src/transformers/models/diffllama/modeling_diffllama.py
fit to changeing "num_heads // 2" place
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* Update src/transformers/models/diffllama/modeling_diffllama.py
fix 2times divide by sqrt(self.head_dim)
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* Update src/transformers/models/diffllama/modeling_diffllama.py
fix 2times divide by sqrt(self.head_dim)
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* Update src/transformers/models/diffllama/modeling_diffllama.py
fit to changeing "num_heads // 2" place.
and more visible
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* I found Attention missed implemented from paper still on e072544a3b.
* re-implemented
* adding groupnorm
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* align with transformers code style
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* fix typo
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* adding groupnorm
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* change SdpaAttention to DiffSdpaAttention
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* fix bug
* Update src/transformers/models/diffllama/modeling_diffllama.py
resolve "not same outputs" problem
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* fix bugs of places of "GroupNorm with scale" and etc
* Revert "fix bugs of places of "GroupNorm with scale" and etc"
This reverts commit 26307d92f6.
* simplify multiple of attention (matmul) operations into one by repeating value_states
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* simplify multiple of attention (matmul) operations into one by repeating value_states
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* simplify multiple of attention (matmul) operations into one by repeating value_states
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
* remove missed type
* add diffllama model_doc
* apply make style/quality
* apply review comment about model
* apply review comment about test
* place diffllama alphabetically on the src/transformers/__init__.py
* fix forgot code
* Supports parameters that are not initialized with standard deviation 0 in the conventional method
* add DiffLlamaConfig to CONFIG_CLASSES_TO_IGNORE_FOR_DOCSTRING_CHECKPOINT_CHECK on utils/check_config_docstrings.py
* remove unused property of config
* add to supported model list
* add to spda supported model list
* fix copyright, remove pretraining_tensor_parallel, and modify for initialization test
* remove unused import and etc.
* empty commit
* empty commit
* empty commit
* apply modular transformers but with bugs
* revert prev commit
* create src/transformers/model/diffllama/modular_diffllama.py
* run utils/modular_model_converter.py
* empty commit
* leaner modular diffllama
* remove more and more in modular_diffllama.pt
* remove more and more in modular_diffllama.pt
* resolve missing docstring entries
* force reset
* convert modular
---------
Co-authored-by: Minho Ryu <ryumin93@gmail.com>
`parallelize()` API is deprecated in favor of accelerate's `device_map="auto"`
and therefore is not accepting new features. At the same time `parallelize()`
implementation is currently CUDA-specific. This commit marks respective
ci tests with `@require_torch_gpu`.
Fixes: #35252
Signed-off-by: Dmitry Rogozhkin <dmitry.v.rogozhkin@intel.com>
* added logic for deleting adapters once loaded
* updated to the latest version of transformers, merged utility function into the source
* updated with missing check
* added peft version check
* Apply suggestions from code review
Co-authored-by: Anton Vlasjuk <73884904+vasqu@users.noreply.github.com>
* changes according to reviewer
* added test for deleting adapter(s)
* styling changes
* styling changes in test
* removed redundant code
* formatted my contributions with ruff
* optimized error handling
* ruff formatted with correct config
* resolved formatting issues
---------
Co-authored-by: Anton Vlasjuk <73884904+vasqu@users.noreply.github.com>
* Make kwargs uniform for SAM
* Remove unused attribute
* Make point_pad_value part of image_kwargs
* Update annotations
* Code review - use existing methods
* Use ProcessorTesterMixin
* Do not add ProcessorTesterMixin everywhere
* fixup mamba2 - caching and several other small fixes
* fixup cached forward
* correct fix this time
* fixup cache - we do not need to extend the attn mask it's handled by generate (gives total ids + mask at each step)
* remove unnecessary (un)squeeze
* fixup cache position
* simplify a few things
* [run-slow] mamba2
* multi gpu attempt two
* [run-slow] mamba2
* [run-slow] mamba2
* [run-slow] mamba2
* [run-slow] mamba2
* add newer slow path fix
* [run-slow] mamba2
* initial cut of modernbert for transformers
* small bug fixes
* fixes
* Update import
* Use compiled mlp->mlp_norm to match research implementation
* Propagate changes in modular to modeling
* Replace duplicate attn_out_dropout in favor of attention_dropout
cc @warner-benjamin let me know if the two should remain separate!
* Update BOS to CLS and EOS to SEP
Please confirm @warner-benjamin
* Set default classifier bias to False, matching research repo
* Update tie_word_embeddings description
* Fix _init_weights for ForMaskedLM
* Match base_model_prefix
* Add compiled_head to match research repo outputs
* Fix imports for ModernBertForMaskedLM
* Just use "gelu" default outright for classifier
* Fix config name typo: initalizer -> initializer
* Remove some unused parameters in docstring. Still lots to edit there!
* Compile the embeddings forward
Not having this resulted in very slight differences - so small it wasn't even noticed for the base model, only for the large model.
But the tiny difference for large propagated at the embedding layer through the rest of the model, leading to notable differences of ~0.0084 average per value, up to 0.2343 for the worst case.
* Add drafts for ForSequenceClassification/ForTokenClassification
* Add initial SDPA support (not exactly equivalent to FA2 yet!)
During testing, FA2 and SDPA still differ by about 0.0098 per value in the token embeddings. It still predicts the correct mask fills, but I'd like to get it fully 1-1 if possible.
* Only use attention dropout if training
* Add initial eager attention support (also not equivalent to FA2 yet!)
Frustratingly, I also can't get eager to be equivalent to FA2 (or sdpa), but it does get really close, i.e. avg ~0.010 difference per value.
Especially if I use fp32 for both FA2&eager, avg ~0.0029 difference per value
The fill-mask results are good with eager.
* Add initial tests, output_attentions, output_hidden_states, prune_heads
Tests are based on BERT, not all tests pass yet: 23 failed, 79 passed, 100 skipped
* Remove kwargs from ModernBertForMaskedLM
Disable sparse_prediction by default to match the normal HF, can be enabled via config
* Remove/adjust/skip improper tests; warn if padding but no attn mask
* Run formatting etc.
* Run python utils/custom_init_isort.py
* FlexAttention with unpadded sequences(matches FA2 within bf16 numerics)
* Reformat init_weights based on review
* self -> module in attention forwards
* Remove if config.tie_word_embeddings
* Reformat output projection on a different line
* Remove pruning
* Remove assert
* Call contiguous() to simplify paths
* Remove prune_qkv_linear_layer
* Format code
* Keep as kwargs, only use if needed
* Remove unused codepaths & related config options
* Remove 3d attn_mask test; fix token classification tuple output
* Reorder: attention_mask above position_ids, fixes gradient checkpointing
* Fix usage if no FA2 or torch v2.5+
* Make torch.compile/triton optional
Should we rename 'compile'? It's a bit vague
* Separate pooling options into separate functions (cls, mean) - cls as default
* Simplify _pad_modernbert_output, remove unused labels path
* Update tied weights to remove decoder.weight, simplify decoder loading
* Adaptively set config.compile based on hf_device_map/device/resize, etc.
* Update ModernBertConfig docstring
* Satisfy some consistency checks, add unfinished docs
* Only set compile to False if there's more than 1 device
* Add docstrings for public ModernBert classes
* Dont replace docstring returns - ends up being duplicate
* Fix mistake in toctree
* Reformat toctree
* Patched FlexAttention, SDPA, Eager with Local Attention
* Implement FA2 -> SDPA -> Eager attn_impl defaulting, crucial
both to match the original performance, and to get the highest inference speed without requiring users to manually pick FA2
* Patch test edge case with Idefics3 not working with 'attn_implementation="sdpa"'
* Repad all_hidden_states as well
* rename config.compile to reference_compile
* disable flex_attention since it crashes
* Update modernbert.md
* Using dtype min to mask in eager
* Fully remove flex attention for now
It's only compatible with the nightly torch 2.6, so we'll leave it be for now. It's also slower than eager/sdpa.
Also, update compile -> reference_compile in one more case
* Call contiguous to allow for .view()
* Copyright 2020 -> 2024
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
* Update/simplify __init__ structure
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
* Remove "... if dropout_prob > 0 else identity"
As dropout with 0.0 should be efficient like identity
* re-use existing pad/unpad functions instead of creating new ones
* remove flexattention method
* Compute attention_mask and local_attention_mask once in modeling
* Simplify sequence classification prediction heads, only CLS now
Users can make custom heads if they feel like it
Also removes the unnecessary pool parameter
* Simplify module.training in eager attn
* Also export ModernBertPreTrainedModel
* Update the documentation with links to finetuning scripts
* Explain local_attention_mask parameter in docstring
* Simplify _autoset_attn_implementation, rely on super()
* Keep "in" to initialize Prediction head
Doublechecked with Benjamin that it's correct/what we used for pretraining
* add back mean pooling
* Use the pooling head in TokenClassification
* update copyright
* Reset config._attn_implementation_internal on failure
* Allow optional attention_mask in ForMaskedLM head
* fix failing run_slow tests
* Add links to the paper
* Remove unpad_no_grad, always pad/unpad without gradients
* local_attention_mask -> sliding_window_mask
* Revert "Use the pooling head in TokenClassification"
This reverts commit 99c38badd1.
There was no real motivation, no info on whether having this bigger head does anything useful.
* Simplify pooling, 2 options via if-else
---------
Co-authored-by: Tom Aarsen <37621491+tomaarsen@users.noreply.github.com>
Co-authored-by: Tom Aarsen <Cubiegamedev@gmail.com>
Co-authored-by: Said Taghadouini <taghadouinisaid@gmail.com>
Co-authored-by: Benjamin Clavié <ben@clavie.eu>
Co-authored-by: Antoine Chaffin <ant54600@hotmail.fr>
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
* initial commit for PR
Co-authored-by: Gabe Goodhart <gabe.l.hart@gmail.com>
* rename dynamic cache
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* add more unit tests
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* add integration test
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* add integration test
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* Add modular bamba file
* Remove trainer changes from unrelated PR
* Modify modular and cofig to get model running
* Fix some CI errors and beam search
* Fix a plethora of bugs from CI/docs/etc
* Add bamba to models with special caches
* Updat to newer mamba PR for mamba sublayer
* fix test_left_padding_compatibility
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* fix style
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* fix remaining tests
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* missed this test
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* ran make style
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* move slow tag to integration obj
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* make style
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* address comments
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* fix modular
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* left out one part of modular
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* change model
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* Make Rotary modular as well
* Update bamba.md
Added overview, update Model inference card and added config
* Update bamba.md
* Update bamba.md
* Update bamba.md
Minor fixes
* Add docs for config and model back
Signed-off-by: Antoni Viros i Martin <aviros@ibm.com>
* Add warning when using fast kernels
* replaced generate example
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
* Address comments from PR
Signed-off-by: Antoni Viros i Martin <aviros@ibm.com>
* Propagate attention fixes
Signed-off-by: Antoni Viros i Martin <aviros@ibm.com>
* Fix attention interfaces to the new API
Signed-off-by: Antoni Viros i Martin <aviros@ibm.com>
* Fix API for decoder layer
Signed-off-by: Antoni Viros i Martin <aviros@ibm.com>
* Remove extra weights
Signed-off-by: Antoni Viros i Martin <aviros@ibm.com>
---------
Signed-off-by: Yu Chin Fabian Lim <flim@sg.ibm.com>
Signed-off-by: Antoni Viros i Martin <aviros@ibm.com>
Co-authored-by: Gabe Goodhart <gabe.l.hart@gmail.com>
Co-authored-by: Antoni Viros i Martin <aviros@ibm.com>
Co-authored-by: divya-kumari32 <72085811+divya-kumari32@users.noreply.github.com>
Co-authored-by: Antoni Viros <ani300@gmail.com>
* do not remove decoder_input_ids for the first segment
* do not remove eos token in generate_with_fallback
* when removing padding tokens, do not remove eos token
* remove eos token in generate (and not in generate_with_fallback!)
* reconciliate short-from/ long-form behavior
* correct avg_logprobs calculation
* handle eos token in segments
* handle decoder_input_ids and eos token in _prepare_decoder_input_ids
* fix incorrect time precision
* always remove eos token
* always remove decoder_input_ids
* no need to handle decoder_inputs_ids and eos token
* no need to remove decoder_input_ids
* no need to handle eos token
* fix num_beams in _retrieve_logit_processors
* remove todo unconsistency
* no need to add eos token
* last_timestamp_pos should indeed be timestamp token pos
* patch generate to enable compatibility with GenerationTesterMixin tests
* adapt test_generate_continue_from_past_key_values
* adapt test_prompt_lookup_decoding_matches_greedy_search
* adapt generic GenerationMixin tests to whisper's generate
* fix speculative decoding
* fix
* [run-slow] whisper
* change HF_HUB_TOKEN for require_read_token
* [run-slow] whisper
* prioritize kwargs over generation_config
* remove unnecessary args
* [run-slow] whisper
* update tests
* [run-slow] whisper
* add comment
* update test
* [run-slow] whisper
* update test + revert require_read_token
* docstring updates
* revert tokenizer decode args change
* do not use a patch + docstring updates
* [run-slow] whisper
* make
* [run-slow] whisper
* add a flag to force unique call to generate
* test update
* [run-slow] whisper
* add force_unique_generate_call arg
* do not use a patch
* correct the timestamps for the pad tokens
* docstring update
* docstring update
* docstring update
* upodate TF tests
* add require_read_token
* [run-slow] whisper
* test reset dynamo
* [run-slow] whisper
* fix
* [run-slow] whisper
* avoid iterating twice on current_segments
* [run-slow] whisper
* [run-slow] whisper
---------
Co-authored-by: Eustache Le Bihan <eustlb@users.noreply.huggingface.co>
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
* feat: add support for sdpa and gradient checkpointing
* fix: ruff format
* fix: config sdpa
* fix: sdpa layer naming convention
* fix: update test_eager_matches_sdpa_inference to handle vision_hidden_states
* test: skip incompatible tests and fix loading issue with sdpa
- Updated tests to skip cases flash and dynamic compile.
- Minor adjustment to ensure correct loading of model with sdpa for dispatch test.
* style: apply Ruff formatting
* ruff fix again after rebase
* [run-slow] sam
* [run-slow] sam
* refactor: Address review comments and improve sub-config handling in SAM model tests
- Added attributes for sub_configs as per PR #34410.
- Enabled tests for configs, ensuring the composite model (SAM) has several sub-configs in the main config.
- Added class attribute _is_composite=True to the tester class
- test_sdpa_can_dispatch_composite_models added
* [run-slow] sam
* style: ruff
* [run-slow] sam
* style: ruff again ...
* [run-slow] sam
* refactor image_processing_auto logic
* fix fast image processor tests
* Fix tests fast vit image processor
* Add safeguard when use_fast True and torchvision not available
* change default use_fast back to None, add warnings
* remove debugging print
* call get_image_processor_class_from_name once
* add more cases
* fix method not found in unittest
Signed-off-by: Lin, Fanli <fanli.lin@intel.com>
* fix more cases
* add more models
* add all
* no unittest.case
* remove for oneformer
* fix style
---------
Signed-off-by: Lin, Fanli <fanli.lin@intel.com>
* draft, run model as compreszed/uncompressed mode
* draft
* run run_compressed=False
* run_compressed as attr
* set run_compressed=False using quantization_config
* remove redundant line
* make is_qat_trainable dependent on run_compressed status
* add tests
* lint
* full in docstring
* add decompress
* comments
* decompress if model is compresssed and not run_compressed
* apply_quant_config logic fix -- populate statedict properly
* comments
* remove non compressed model
* make is_compressed as property
* cosmetic
* run apply_quant_config for non-compressed models -- popualte scales and zeropoints
* add pahtway for decompressing sparse models
* typo on is_quantization_compressed
* lint
* fix typo
