Commit Graph

4407 Commits

Author SHA1 Message Date
Pavel Iakubovskii
099d93d2e9
Grounding DINO Processor standardization (#34853)
* Add input ids to model output

* Add text preprocessing for processor

* Fix snippet

* Add test for equivalence

* Add type checking guard

* Fixing typehint

* Fix test for added `input_ids` in output

* Add deprecations and "text_labels" to output

* Adjust tests

* Fix test

* Update code examples

* Minor docs and code improvement

* Remove one-liner functions and rename class to CamelCase

* Update docstring

* Fixup
2025-01-17 14:18:16 +00:00
Pavel Iakubovskii
42b2857b01
OmDet Turbo processor standardization (#34937)
* Fix docstring

* Fix docstring

* Add `classes_structure` to model output

* Update omdet postprocessing

* Adjust tests

* Update code example in docs

* Add deprecation to "classes" key in output

* Types, docs

* Fixing test

* Fix missed clip_boxes

* [run-slow] omdet_turbo

* Apply suggestions from code review

Co-authored-by: Yoni Gozlan <74535834+yonigozlan@users.noreply.github.com>

* Make CamelCase class

---------

Co-authored-by: Yoni Gozlan <74535834+yonigozlan@users.noreply.github.com>
2025-01-17 14:10:19 +00:00
Pavel Iakubovskii
94ae9a8da1
OwlViT/Owlv2 post processing standardization (#34929)
* Refactor owlvit post_process_object_detection + add text_labels

* Fix copies in grounding dino

* Sync with Owlv2 postprocessing

* Add post_process_grounded_object_detection method to processor, deprecate post_process_object_detection

* Add test cases

* Move text_labels to processors only

* [run-slow] owlvit owlv2

* [run-slow] owlvit, owlv2

* Update snippets

* Update docs structure

* Update deprecated objects for check_repo

* Update docstring for post processing of image guided object detection
2025-01-17 13:58:28 +00:00
Cyril Vallez
ab1afd56f5
Fix some tests (#35682)
* cohere tests

* glm tests

* cohere2 model name

* create decorator

* update

* fix cohere2 completions

* style

* style

* style

* add cuda in comments
2025-01-17 12:10:43 +00:00
Ross Wightman
8c1b5d3782
🚨🚨🚨 An attempt to fix #29554. Include 'LayerNorm.' in gamma/beta rename scope, optimize string search. (#35615)
* An attempt to fix #29554. Include 'LayerNorm.' in gamma/beta rename scope, reduce number of characters searched on every load considerably.

* Fix fix on load issue

* Fix gamma/beta warning test

* A style complaint

* Improve efficiency of weight norm key rename. Add better comments about weight norm and layer norm renaming.

* Habitual elif redunant with the return
2025-01-16 17:25:44 -08:00
Joao Gante
94af1c0aa2
[generate] return Cache object even if passed in a legacy format (#35673)
* generate returns a Cache object by default

* fix tests

* fix test for encoder-decoder models
2025-01-16 17:06:24 +00:00
Joao Gante
2818307e93
[generate] can instantiate GenerationConfig(cache_implementation="static") (#35679)
fix failing instantiation
2025-01-16 17:04:54 +00:00
Joao Gante
aeeceb9916
[cache] add a test to confirm we can use cache at train time (#35709)
* add test

* augment test as suggested

* Update tests/utils/test_modeling_utils.py

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

* rerun tests

---------

Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
2025-01-16 17:02:34 +00:00
kang sheng
2cbcc5877d
Fix condition when GA loss bug fix is not performed (#35651)
* fix condition when GA loss bug fix is not performed

* max loss diff is 2.29

* fix typo

* add an extra validation that loss should not vary too much
2025-01-16 13:59:53 +01:00
jiqing-feng
387663e571
Enable gptqmodel (#35012)
* gptqmodel

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* fix format

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* update readme

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* gptqmodel need use checkpoint_format (#1)

* gptqmodel need use checkpoint_format

* fix quantize

* Update quantization_config.py

* Update quantization_config.py

* Update quantization_config.py

---------

Co-authored-by: ZX-ModelCloud <zx@modelcloud.ai>
Co-authored-by: Qubitium-ModelCloud <qubitium@modelcloud.ai>

* Revert quantizer_gptq.py (#2)

* revert quantizer_gptq.py change

* pass **kwargs

* limit gptqmodel and optimum version

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* fix format

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* fix warning

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* fix version check

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* revert unrelated changes

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* enable gptqmodel tests

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* fix requires gptq

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* Fix Transformer compat (#3)

* revert quantizer_gptq.py change

* pass **kwargs

* add meta info

* cleanup

* cleanup

* Update quantization_config.py

* hf_select_quant_linear pass checkpoint_format and meta

* fix GPTQTestCUDA

* Update test_gptq.py

* gptqmodel.hf_select_quant_linear() now does not select ExllamaV2

* cleanup

* add backend

* cleanup

* cleanup

* no need check exllama version

* Update quantization_config.py

* lower checkpoint_format and backend

* check none

* cleanup

* Update quantization_config.py

* fix self.use_exllama == False

* spell

* fix unittest

* fix unittest

---------

Co-authored-by: LRL <lrl@lbx.dev>
Co-authored-by: Qubitium-ModelCloud <qubitium@modelcloud.ai>

* fix format

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* fix format again

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* update gptqmodel version (#6)

* update gptqmodel version

* update gptqmodel version

* fix unit test (#5)

* update gptqmodel version

* update gptqmodel version

* "not self.use_exllama" is not equivalent to "self.use_exllama==False"

* fix unittest

* update gptqmodel version

* backend is loading_attibutes (#7)

* fix format and tests

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* fix memory check

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* fix device mismatch

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* fix result check

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* Update src/transformers/quantizers/quantizer_gptq.py

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

* Update src/transformers/quantizers/quantizer_gptq.py

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

* Update src/transformers/quantizers/quantizer_gptq.py

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

* update tests

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* review: update docs (#10)

* review: update docs (#12)

* review: update docs

* fix typo

* update tests for gptqmodel

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* update document (#9)

* update overview.md

* cleanup

* Update overview.md

* Update overview.md

* Update overview.md

* update gptq.md

* Update gptq.md

* Update gptq.md

* Update gptq.md

* Update gptq.md

* Update gptq.md

* Update gptq.md

---------

Co-authored-by: Qubitium-ModelCloud <qubitium@modelcloud.ai>

* typo

* doc note for asymmetric quant

* typo with apple silicon(e)

* typo for marlin

* column name revert: review

* doc rocm support

* Update docs/source/en/quantization/gptq.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/quantization/gptq.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/quantization/gptq.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/quantization/gptq.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/quantization/overview.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/quantization/overview.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

---------

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>
Co-authored-by: LRL-ModelCloud <165116337+LRL-ModelCloud@users.noreply.github.com>
Co-authored-by: ZX-ModelCloud <zx@modelcloud.ai>
Co-authored-by: Qubitium-ModelCloud <qubitium@modelcloud.ai>
Co-authored-by: ZX-ModelCloud <165115237+ZX-ModelCloud@users.noreply.github.com>
Co-authored-by: LRL <lrl@lbx.dev>
Co-authored-by: Marc Sun <57196510+SunMarc@users.noreply.github.com>
Co-authored-by: Mohamed Mekkouri <93391238+MekkCyber@users.noreply.github.com>
Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>
2025-01-15 14:22:49 +01:00
Raushan Turganbay
09d5f76274
Clean-up composite configs (#34603)
* remove manual assignment tie-word-embeddings

* remove another unused attribute

* fix tests

* fix tests

* remove unnecessary overwrites

* fix

* decoder=True

* clean pix2struct

* run-all

* forgot `_tied_weights_keys` when adding Emu3

* also Aria + fix-copies

* and clean aria
2025-01-15 10:04:07 +01:00
Mahdi Baghbanzadeh
c61fcde910
Enhance DataCollatorForLanguageModeling with Configurable Token Replacement Probabilities (#35251)
* DataCollatorForLanguageModeling class was updated with new parameters that provides more control over the token masking and relacing

* DataCollatorForLanguageModeling class was updated with new parameters that provides more control over the token masking and relacing

* Addressed review comments, modified the docstring and made a test for the DataCollatorForLanguageModeling
2025-01-14 17:01:10 +00:00
Mohamed Mekkouri
a11041ffad
Fix : add require_read_token for gemma2 gated model (#35687)
fix gemma2 gated model test
2025-01-14 11:47:05 +01:00
Mohamed Mekkouri
df2a812e95
Fix expected output for ggml test (#35686)
fix expected output
2025-01-14 11:46:55 +01:00
Mohamed Mekkouri
050636518a
Fix : HQQ config when hqq not available (#35655)
* fix

* make style

* adding require_hqq

* make style
2025-01-14 11:37:37 +01:00
Arthur
c23a1c1932
Add-helium (#35669)
* Add the helium model.

* Add a missing helium.

* And add another missing helium.

* Use float for the rmsnorm mul.

* Add the Helium tokenizer converter.

* Add the pad token as suggested by Arthur.

* Update the RMSNorm + some other tweaks.

* Fix more rebase issues.

* fix copies and style

* fixes and add helium.md

* add missing tests

* udpate the backlink

* oups

* style

* update init, and expected results

* small fixes

* match test outputs

* style fixup, fix doc builder

* add dummies and we should be good to go!z

* update sdpa and fa2 documentation

---------

Co-authored-by: laurent <laurent.mazare@gmail.com>
2025-01-13 18:41:15 +01:00
Fanli Lin
2fa876d2d8
[tests] make cuda-only tests device-agnostic (#35607)
* intial commit

* remove unrelated files

* further remove

* Update test_trainer.py

* fix style
2025-01-13 14:48:39 +01:00
Arthur
e6f9b03464
[Compile] Only test compiling model forward pass (#35658)
* rename test to only compile forward!

* style emu
2025-01-13 13:43:29 +01:00
Raushan Turganbay
84a6789145
Enable different torch dtype in sub models (#34873)
* fix

* fix test

* add tests

* add more tests

* fix tests

* supposed to be a torch.dtype test

* handle BC and make fp32 default
2025-01-13 13:42:08 +01:00
Yih-Dar
1e3c6c1f7d
Skip MobileNetV1ModelTest::test_batching_equivalence for now (#35614)
fix

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
2025-01-10 18:32:36 +01:00
Yih-Dar
04eae987f3
Fix flaky test_beam_search_low_memory (#35611)
* fix

* fix

* fix

* fix

* fix

---------

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
2025-01-10 17:31:03 +01:00
Zach Mueller
b02828e4af
Let EarlyStoppingCallback not require load_best_model_at_end (#35101)
* Bookmark

* Add warning
2025-01-10 10:25:32 -05:00
Zach Mueller
1211e616a4
Use inherit tempdir makers for tests + fix failing DS tests (#35600)
* Use existing APIs to make tempdir folders

* Fixup deepspeed too

* output_dir -> tmp_dir
2025-01-10 10:01:58 -05:00
Yih-Dar
bbc00046b9
Fix flaky test_custom_4d_attention_mask (#35606)
* fix

* fix

---------

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
2025-01-10 15:40:04 +01:00
Raushan Turganbay
52e1f87c7d
[WIP] Emu3: add model (#33770)
* model can convert to HF and be loaded back

* nit

* works in single batch generation but hallucinates

* use the image tokens

* add image generation

* now it works

* add tests

* update

* add modulare but it doesn't work for porting docstring :(

* skip some tests

* add slow tests

* modular removed the import?

* guess this works

* update

* update

* fix copies

* fix test

* fix copies

* update

* docs

* fix tests

* last fix tests?

* pls

* repo consistency

* more style

* style

* remove file

* address comments

* tiny bits

* update after the new modular

* fix tests

* add one more cond in check attributes

* decompose down/up/mid blocks

* allow static cache generation in VLMs

* nit

* fix copies

* Update docs/source/en/model_doc/emu3.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/model_doc/emu3.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/model_doc/emu3.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/model_doc/emu3.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/model_doc/emu3.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/model_doc/emu3.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/model_doc/emu3.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/model_doc/emu3.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* fix VAE upsampling

* Update src/transformers/models/emu3/modular_emu3.py

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

* address comments

* state overwritten stuff explicitly

* fix copies

* add the flag for flex attn

---------

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
2025-01-10 12:23:00 +01:00
Cyril Vallez
ccc0381d36
Fix flex_attention in training mode (#35605)
* fix flex

* add test

* style
2025-01-10 11:49:12 +01:00
Raushan Turganbay
e0646f3dce
Chat template: return vectorized output in processors (#34275)
* update chat template

* style

* fix tests

* Update src/transformers/image_utils.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* typehints + docs

* fix tests

* remove unnecessary warnings

* forgot code style :(

* allow users to pass backend and num frames

* Update docs/source/en/chat_templating.md

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update src/transformers/image_utils.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update src/transformers/image_utils.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update src/transformers/image_utils.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update src/transformers/image_utils.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update src/transformers/image_utils.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update src/transformers/image_utils.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update src/transformers/processing_utils.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* typo fix

* style

* address comments

* align with "pipeline" template

* update docs

* update docs

* unpack for all kwargs?

* wrong conflict resolution while rebasing

* tmp

* update docs

* Update docs/source/en/chat_templating.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/chat_templating.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/chat_templating.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

* Update docs/source/en/chat_templating.md

Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>

---------

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>
Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>
2025-01-10 11:05:29 +01:00
eustlb
5f087d1335
Add Moonshine (#34784)
* config draft

* full encoder forward

* full decoder forward

* fix sdpa and FA2

* fix sdpa and FA2

* moonshine model

* moonshine model forward

* fix attention with past_key_values

* add MoonshineForConditionalGeneration

* fix cache handling and causality for cross attention

* no causal attention mask for the encoder

* model addition (imports etc)

* small nit

* nits

* Update src/transformers/models/moonshine/convert_usefulsensors_to_hf.py

Co-authored-by: Joshua Lochner <admin@xenova.com>

* add rope_theta

* nits

* model doc

* Update src/transformers/models/auto/configuration_auto.py

Co-authored-by: Joshua Lochner <admin@xenova.com>

* imports

* add MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES

* updates modular

* make

* make fix-copies

* ruff check examples fix

* fix check_modular_conversion

* nit

* nits

* nits

* copied from -> imports

* imports fix

* integrate attention refacto

* modular edge case

* remove encoder

* convolutions params in config

* run modular_model_converter

* make

* Update docs/source/en/model_doc/moonshine.md

Co-authored-by: Joshua Lochner <admin@xenova.com>

* MoonshineModelTest

* correct typo

* make style

* integration tests

* make

* modular convert

* name conversion update (up_proj -> fc1 etc)

* update config

* update MLP

* update attention

* update encoder layer

* update decoder layer

* update convolutions parameters

* update encoder

* remove INPUTS_DOCSTRING

* update decoder

* update conditional generation

* update pretrained model

* imports

* modular converted

* update doc

* fix

* typo

* update doc

* update license

* update init

* split config in file

* two classes for MLP

* attention from GLM

* from GlmRotaryEmbedding

* split MLP

* apply arthur's review suggestions

* apply arthur's review suggestions

* apply arthur's review suggestions

* auto feature extractor

* convert modular

* fix + make

* convert modular

* make

* unsplit config

* use correct checkpoint

* wrap generate

* update tests

* typos

* make

* typo

* update doc

---------

Co-authored-by: Joshua Lochner <admin@xenova.com>
2025-01-10 11:00:54 +01:00
Yih-Dar
6f127d3f81
Skip torchscript tests if a cache object is in model's outputs (#35596)
* fix 1

* fix 1

* comment

---------

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
2025-01-10 10:46:03 +01:00
Tom Aarsen
6b73ee8905
ModernBert: reuse GemmaRotaryEmbedding via modular + Integration tests (#35459)
* Introduce 5 integration tests for the 4 model classes + torch export

* ModernBert: reuse GemmaRotaryEmbedding via modular

* Revert #35589, keep rope_kwargs; rely on them in modular_modernbert

* Revert "Revert #35589, keep rope_kwargs; rely on them in modular_modernbert"

This reverts commit 11b44b9ee8.

* Don't set rope_kwargs; override 'self.rope_init_fn' call instead
2025-01-10 10:25:10 +01:00
Cyril Vallez
3a4ae6eace
Refactor/fix Cohere2 (#35594)
* refactor/fix cohere2

* add kwargs

* tests

* remove func and import it
2025-01-09 17:54:57 +01:00
Tom Aarsen
32e0db8a69
[tokenizers] Ensure that add_prefix_space is propagated to backend_tokenizer.pre_tokenizer (#35593)
* Ensure that add_prefix_space is propagated to backend_tokenizer.pre_tokenizer

in PreTrainedTokenizerFast, rather than relying on subclasses to take care of this.

* Simplify setting self.add_prefix_space, ensure pre_tok exists

* Wrap in try-except to catch 'Custom PreTokenizer cannot be serialized'

862d1a346a/bindings/python/src/pre_tokenizers.rs (L672) produces the Exception. They're triggered by the roformer tests, as the RoFormerTokenizerFast uses a custom PreTokenizer.

* Propagate add_prefix_space in T5TokenizerFast to superclass
2025-01-09 17:46:50 +01:00
Cyril Vallez
46276f9a7f
Fix modular edge case + modular sorting order (#35562)
* look-ahead negation

* re add examples by default

* Fix the bug in topological sort

* Update create_dependency_mapping.py

* start adding test

* finalize test

* more tests

* style

* style
2025-01-09 17:17:52 +01:00
Yih-Dar
82dd6c14bb
Fix flaky SwitchTransformersModelTest::test_training_gradient (#35587)
* fix

* Update tests/models/switch_transformers/test_modeling_switch_transformers.py

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

---------

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
2025-01-09 15:36:22 +01:00
Arthur
eb4579cf43
tokenizer train from iterator without pre_tokenizers (#35396)
* fix if else issues

* add a test

* fix the test

* style
2025-01-09 15:34:43 +01:00
Jack Morris
832c6191ed
Add inputs_embeds param to ModernBertModel (#35373)
* update modular_modernbert -- add inputs_embeds param to ModernBertModel

* Fix implementation issues; extend to other classes; docstring

First of all, the inputs_embeds shouldn't fully replace `self.embeddings(input_ids)`, because this call also does layer normalization and dropout. So, now both input_ids and inputs_embeds is passed to the ModernBertEmbeddings, much like how BertEmbeddings is implemented.

I also added `inputs_embeds` to the docstring, and propagated the changes to the other model classes.

I also introduced an error if input_ids and input_embeds are both or neither provided.

Lastly, I fixed an issue with device being based solely on input_ids with attention_mask.

* Propagate inputs_embeds to ModernBertForMaskedLM correctly

Also reintroduce inputs_embeds test

---------

Co-authored-by: Tom Aarsen <Cubiegamedev@gmail.com>
2025-01-09 14:17:26 +01:00
Yih-Dar
1b2f942af7
Fix flaky test_batching_equivalence (#35564)
* yes!

* oh no!!!

* oh no!!!

* style

* oh no!!!

* oh no!!!

* oh no!!!

* oh no!!!

---------

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
2025-01-09 14:00:08 +01:00
Cyril Vallez
965a2fb320
More model refactoring! (#35359)
* cohere

* style

* phi3

* style

* small fix

* small fix

* phi3 longrope

* oups

* Update rope (only for phi3 still)

* Update test_modeling_rope_utils.py

* Update modeling_phi3.py

* fix

* fix copies

* style

* Fix copied from bad renaming
2025-01-09 11:09:09 +01:00
nhamanasu
b32938aeee
Fix all output_dir in test_trainer.py to use tmp_dir (#35266)
* update codecarbon

* replace directly-specified-test-dirs with tmp_dir

* pass tmp_dir to all get_regression_trainer

* test_trainer.py: Use tmp_dir consistently for all output_dir arguments

* fix some with...as tmp_dir blocks

* reflect the comments to improve test_trainer.py

* refresh .gitignore
2025-01-08 19:44:39 +01:00
Joao Gante
76da6ca034
Pipeline: simple API for assisted generation (#34504)
Co-authored-by: Matt <Rocketknight1@users.noreply.github.com>
2025-01-08 17:08:02 +00:00
Arthur
3f483beab9
[PixtralLarge] Update Pixtral conversion script to support large format! (#34801)
* update conversion script

* update for bias again

* remove pdv

* use my dir

* Update how we initialize the tokenizer

* Convert in bfloat16

* Undo that one again

* fix config dump

* .to() was broken for BatchMixFeature

* quick debug breakpoint

* put the breakpoint in the right place

* Add a config flag for the multimodal projector bias

* Add a config flag for the multimodal projector bias

* Conversion script can load chat templates

* Indent config for comparison

* Stop clobbering the config

* Re-enable the config clobber

* Get rid of the config manual save - it has no effect!

* Handle adapter bias correctly

* Default vision transformer activation to silu

* Remove legacy processing path

* One commit with all the debug breakpoints before I delete them all, in case I need to revert

* Update conversion

* Remove vLLM debugging instrumentation

* Drop xformers

* Remove debug enumerates

* make fixup

* make fixup

* Break copied from in pixtral

* Propagate multimodal_projector_bias change

* Propagate multimodal_projector_bias change

* Remove debug device .to()

* Restore attention weights output

* Fix Pixtral test

* Drop image_seq_length

* Drop image_seq_length

* Put the legacy processing code back

* Add the bias option to the llava_next_video config

* Add the bias option to the llava_next_video config

* Make certain args required in converter

* Make certain args required in converter

* typo

* make fixup

* Reverting some dtype changes since it seems to work without them

---------

Co-authored-by: arthur@huggingface.co <arthur@ip-26-0-166-244.ec2.internal>
Co-authored-by: Matt <rocketknight1@gmail.com>
Co-authored-by: Matt <Rocketknight1@users.noreply.github.com>
2025-01-08 17:39:47 +01:00
NielsRogge
8490d3159c
Add ViTPose (#30530)
* First draft

* Make fixup

* Make forward pass worké

* Improve code

* More improvements

* More improvements

* Make predictions match

* More improvements

* Improve image processor

* Fix model tests

* Add classic decoder

* Convert classic decoder

* Verify image processor

* Fix classic decoder logits

* Clean up

* Add post_process_pose_estimation

* Improve post_process_pose_estimation

* Use AutoBackbone

* Add support for MoE models

* Fix tests, improve num_experts%

* Improve variable names

* Make fixup

* More improvements

* Improve post_process_pose_estimation

* Compute centers and scales

* Improve postprocessing

* More improvements

* Fix ViTPoseBackbone tests

* Add docstrings, fix image processor tests

* Update index

* Use is_cv2_available

* Add model to toctree

* Add cv2 to doc tests

* Remove script

* Improve conversion script

* Add coco_to_pascal_voc

* Add box_to_center_and_scale to image_transforms

* Update tests

* Add integration test

* Fix merge

* Address comments

* Replace numpy by pytorch, improve docstrings

* Remove get_input_embeddings

* Address comments

* Move coco_to_pascal_voc

* Address comment

* Fix style

* Address comments

* Fix test

* Address comment

* Remove udp

* Remove comment

* [WIP] need to check if the numpy function is same as cv

* add scipy affine_transform

* Update src/transformers/models/vitpose/image_processing_vitpose.py

Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* refactor convert

* add output_shape

* add atol 5e-2

* Use hf_hub_download in conversion script

* make box_to_center more applicable

* skipt test_get_set_embedding

* fix to accept array and fix CI

* add co-contributor

* make it to tensor type output

* add torch

* change to torch tensor

* add more test

* minor change

* CI test change

* import torch should be above ImageProcessor

* make style

* try not use torch in def

* Update src/transformers/models/vitpose/image_processing_vitpose.py

Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/vitpose_backbone/configuration_vitpose_backbone.py

Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/vitpose_backbone/modeling_vitpose_backbone.py

Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/vitpose/modeling_vitpose.py

Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* fix

* fix

* add caution

* make more detail about dataset_index

* Update src/transformers/models/vitpose/modeling_vitpose.py

Co-authored-by: Sangbum Daniel Choi <34004152+SangbumChoi@users.noreply.github.com>

* Update src/transformers/models/vitpose/image_processing_vitpose.py

Co-authored-by: Sangbum Daniel Choi <34004152+SangbumChoi@users.noreply.github.com>

* add docs

* Update docs/source/en/model_doc/vitpose.md

* Update src/transformers/models/vitpose/configuration_vitpose.py

Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/__init__.py

Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Revert "Update src/transformers/__init__.py"

This reverts commit 7ffa504450.

* change name

* Update src/transformers/models/vitpose/image_processing_vitpose.py

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

* Update tests/models/vitpose/test_modeling_vitpose.py

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

* Update docs/source/en/model_doc/vitpose.md

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

* Update src/transformers/models/vitpose/modeling_vitpose.py

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

* Update src/transformers/models/vitpose_backbone/modeling_vitpose_backbone.py

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

* Update src/transformers/models/vitpose/image_processing_vitpose.py

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

* move vitpose only function to image_processor

* raise valueerror when using timm backbone

* use out_indices

* Update src/transformers/models/vitpose/image_processing_vitpose.py

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

* remove camel-case of def flip_back

* rename vitposeEstimatorOutput

* Update src/transformers/models/vitpose_backbone/modeling_vitpose_backbone.py

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

* fix confused camelcase of MLP

* remove in-place logic

* clear scale description

* make consistent batch format

* docs update

* formatting docstring

* add batch tests

* test docs change

* Update src/transformers/models/vitpose/image_processing_vitpose.py

* Update src/transformers/models/vitpose/configuration_vitpose.py

* chagne ViT to Vit

* change to enable MoE

* make fix-copies

* Update docs/source/en/model_doc/vitpose.md

Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* extract udp

* add more described docs

* simple fix

* change to accept target_size

* make style

* Update src/transformers/models/vitpose/image_processing_vitpose.py

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

* Update src/transformers/models/vitpose/configuration_vitpose.py

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

* change to `verify_backbone_config_arguments`

* Update docs/source/en/model_doc/vitpose.md

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

* remove unnecessary copy

* make config immutable

* enable gradient checkpointing

* update inappropriate docstring

* linting docs

* split function for visibility

* make style

* check isinstances

* change to acceptable use_pretrained_backbone

* make style

* remove copy in docs

* Update src/transformers/models/vitpose_backbone/modeling_vitpose_backbone.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update docs/source/en/model_doc/vitpose.md

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update src/transformers/models/vitpose/modeling_vitpose.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* simple fix + make style

* change input config of activation function to string

* Update docs/source/en/model_doc/vitpose.md

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* tmp docs

* delete index.md

* make fix-copies

* simple fix

* change conversion to sam2/mllama style

* Update src/transformers/models/vitpose/image_processing_vitpose.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update src/transformers/models/vitpose/image_processing_vitpose.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* refactor convert

* add supervision

* Update src/transformers/models/vitpose_backbone/modeling_vitpose_backbone.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* remove reduntant def

* seperate code block for visualization

* add validation for num_moe

* final commit

* add labels

* [run-slow] vitpose, vitpose_backbone

* Update src/transformers/models/vitpose/convert_vitpose_to_hf.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* enable all conversion

* final commit

* [run-slow] vitpose, vitpose_backbone

* ruff check --fix

* [run-slow] vitpose, vitpose_backbone

* rename split module

* [run-slow] vitpose, vitpose_backbone

* fix pos_embed

* Simplify init

* Revert "fix pos_embed"

This reverts commit 2c56a4806e.

* refactor single loop

* allow flag to enable custom model

* efficiency of MoE to not use unused experts

* make style

* Fix range -> arange to avoid warning

* Revert MOE router, a new one does not work

* Fix postprocessing a bit (labels)

* Fix type hint

* Fix docs snippets

* Fix links to checkpoints

* Fix checkpoints in tests

* Fix test

* Add image to docs

---------

Co-authored-by: Niels Rogge <nielsrogge@nielss-mbp.home>
Co-authored-by: Niels Rogge <nielsrogge@Nielss-MacBook-Pro.local>
Co-authored-by: sangbumchoi <danielsejong55@gmail.com>
Co-authored-by: Sangbum Daniel Choi <34004152+SangbumChoi@users.noreply.github.com>
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>
Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>
2025-01-08 16:02:14 +00:00
Minho Shim
4349a0e401
fix: Qwen2-VL generate with inputs_embeds (#35466)
* fix: Qwen2-VL generate with inputs_embeds

* change: optional input_ids in get_rope_index
2025-01-08 16:36:03 +01:00
Sean (Seok-Won) Yi
88e18b3c63
Update doc for metric_for_best_model when save_strategy="best". (#35389)
* 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.
2025-01-08 16:32:35 +01:00
Pavel Iakubovskii
657bb14f98
Enable auto task for timm models in pipeline (#35531)
* Enable auto task for timm models

* Add pipeline test
2025-01-08 15:14:17 +00:00
Pavel Iakubovskii
59e5b3f01b
Timm wrapper label names (#35553)
* Add timm wrapper label names mapping

* Add index to classification pipeline

* Revert adding index for pipelines

* Add custom model check for loading timm labels

* Add tests for labels

* [run-slow] timm_wrapper

* Add note regarding label2id mapping
2025-01-08 14:09:46 +00:00
Jacky Lee
3c1895aa65
Fix Qwen2VL processor to handle odd number of frames (#35431)
* 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
2025-01-08 13:49:00 +01:00
Quentin Lhoest
3fde88b19d
support chat generator as input of TextGenerationPipeline (#35551)
* support chat generator as input of TextGenerationPipeline

* missing import

* fix tests

* again

* simpler

* add test
2025-01-08 13:27:07 +01:00
Raushan Turganbay
d1681ec2b6
VLMs: major clean up 🧼 (#34502)
only lllava models are modified
2025-01-08 10:35:23 +01:00
Jade Choghari
7176e06b52
Add TextNet (#34979)
* WIP

* Add config and modeling for Fast model

* Refactor modeling and add tests

* More changes

* WIP

* Add tests

* Add conversion script

* Add conversion scripts, integration tests, image processor

* Fix style and copies

* Add fast model to init

* Add fast model in docs and other places

* Fix import of cv2

* Rename image processing method

* Fix build

* Fix Build

* fix style and fix copies

* Fix build

* Fix build

* Fix Build

* Clean up docstrings

* Fix Build

* Fix Build

* Fix Build

* Fix build

* Add test for image_processing_fast and add documentation tests

* some refactorings

* Fix failing tests

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Introduce TextNet

* Fix failures

* Refactor textnet model

* Fix failures

* Add cv2 to setup

* Fix failures

* Fix failures

* Add CV2 dependency

* Fix bugs

* Fix build issue

* Fix failures

* Remove textnet from modeling fast

* Fix build and other things

* Fix build

* some cleanups

* some cleanups

* Some more cleanups

* Fix build

* Incorporate PR feedbacks

* More cleanup

* More cleanup

* More cleanup

* Fix build

* Remove all the references of fast model

* More cleanup

* Fix build

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Fix Build

* Fix build

* Fix build

* Fix build

* Fix build

* Fix build

* Incorporate PR feedbacks

* Fix style

* Fix build

* Incorporate PR feedbacks

* Fix image processing mean and std

* Incorporate PR feedbacks

* fix build failure

* Add assertion to image processor

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* fix style failures

* fix build

* Fix Imageclassification's linear layer, also introduce TextNetImageProcessor

* Fix build

* Fix build

* Fix build

* Fix build

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Fix build

* Incorporate PR feedbacks

* Remove some script

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Incorporate PR feedbacks

* Fix image processing in textnet

* Incorporate PR Feedbacks

* Fix CI failures

* Fix failing test

* Fix failing test

* Fix failing test

* Fix failing test

* Fix failing test

* Fix failing test

* Add textnet to readme

* Improve readability

* Incorporate PR feedbacks

* fix code style

* fix key error and convert working

* tvlt shouldn't be here

* fix test modeling test

* Fix tests, make fixup

* Make fixup

* Make fixup

* Remove TEXTNET_PRETRAINED_MODEL_ARCHIVE_LIST

* improve type annotation

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update tests/models/textnet/test_image_processing_textnet.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* improve type annotation

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* space typo

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* improve type annotation

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update src/transformers/models/textnet/configuration_textnet.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* make conv layer kernel sizes and strides default to None

* Update src/transformers/models/textnet/modeling_textnet.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update src/transformers/models/textnet/modeling_textnet.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* fix keyword bug

* add batch init and make fixup

* Make fixup

* Update integration test

* Add figure

* Update textnet.md

* add testing and fix errors (classification, imgprocess)

* fix error check

* make fixup

* make fixup

* revert to original docstring

* add make style

* remove conflict for now

* Update modeling_auto.py

got a confusion in `timm_wrapper` - was giving some conflicts

* Update tests/models/textnet/test_modeling_textnet.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update src/transformers/models/textnet/modeling_textnet.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update tests/models/textnet/test_modeling_textnet.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* Update src/transformers/models/textnet/modeling_textnet.py

Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>

* add changes

* Update textnet.md

* add doc

* add authors hf ckpt + rename

* add feedback: classifier/docs

---------

Co-authored-by: raghavanone <opensourcemaniacfreak@gmail.com>
Co-authored-by: jadechoghari <jadechoghari@users.noreply.huggingface.co>
Co-authored-by: Niels <niels.rogge1@gmail.com>
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>
Co-authored-by: Pavel Iakubovskii <qubvel@gmail.com>
2025-01-08 09:52:51 +01:00