* Add cohere2_vision to support CohereLabs/command-a-vision-07-2025
* update and add modualr file
* update processors and check with orig impl later
* delete unused files
* image processor reduce LOC and re-use GotOCR2
* update the config to use modular
* model tests pass
* processor fixes
* check model outputs decorator
* address one more comment
* Update tokens. Temp - need to read from tokenizer'
* fix for multi-gpu
* Fix image token handling
* upadte image token expansion logic
* fix a few issues with remote code loading
* not related but modular forces us to change all files now
* Add overview and code sample to cohere vision docs
* add scripts. TMP.
* Update inference script
* Create script
* set dtype in export script
* TO revert: modular export fix
* Fix scripts
* Revert "TO revert: modular export fix"
This reverts commit bdb2f305b61027a05f0032ce70d6ca698879191c.
* Use modular weights
* Upload to hub
Removed OOD weights ad script
* Updated docs
* fix import error
Update docs
Added pipeline test
* Updated docs
* Run modular script
remove modular for config
Added patch_size
Added docstrings in modular
Fix OOM
Add docs, fixup integration tests. 8-gpu passing
* tiny updates
* address comments + fixup
* add test for chat template
* check model outputs workaround
* aya vision fix check model inputs
* Revert "add test for chat template"
This reverts commit 42c756e397f588d76b449ff1f93292d8ee0202d8.
* reveert more changes
* last revert
* skip and merge
* faulty copy from
---------
Co-authored-by: Julian Mack <julian.mack@cohere.com>
Co-authored-by: kyle-cohere <kyle@cohere.com>
* Add missing cache_position argument.
* Pass cache_position to language model.
* Overwrite prepare_inputs_for_generation.
* Set model to half precision for Flash Attention test.
* Cast model to bfloat16.
* fix most tests
* skip a few more tests
* address comments
* fix chameleon tests
* forgot to uncomment
* qwen has its own tests with images, rename it as well
* use partial to wrap around `transformers` utils!
* try to refactor?
* revert one wrong change
* just a nit
* push
* reverter watever was wrong!
* some nits
* fixes when there is no attention mask
* bring the licence back
* some fixes
* nit
* style
* remove prints
* correct dtype
* fa flags for testing
* update
* use paged attention if requested!
* updates
* a clone was needed, not sure why
* automatically create cu seq lens when input is flash, this at least makes sure layers don't re-compute
* simplify and improve?
* flash attention is kinda broken on recent cuda version so allow the opportunity to use something else
* fix!
* protect kernels import
* update
* properly parse generation config being passed
* revert and update
* add two tests
* some fixes
* fix test FA2
* takes comment into account
* fixup
* revert changes
* revert the clone, it is only needed because the metal kernel is not doing it?
* [docs] update attention implementation and cache docs (#39547)
* update docs
* Apply suggestions from code review
Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>
* applu suggestions
---------
Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>
* fix mps on our side for now
* Update src/transformers/integrations/flash_paged.py
* no qa
---------
Co-authored-by: Vasqu <antonprogamer@gmail.com>
Co-authored-by: Raushan Turganbay <raushan@huggingface.co>
Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com>
* init
* copied from remote
* add proper structure and llama like structure
* fixup
* revert to state that works
* get closer to llama
* slow and steady
* some removal
* masks work
* it is indeed the rope implementation, how dafuq does it mesh with the cache now hmm
* nice
* getting closer
* closer to transformers style
* let's simplify this, batching works now
* simplified
* working version with modular
* it is indeed the rotation per weights, make it complete llama style
* cleanup conversion, next to look at -> tokenizer
* remove llama artefacts
* fix modeling tests (common ones)
* style
* integration test + first look into tokenization (will need more work, focussing on modeling other models first)
* style
* working moe version, based on remote
* lets keep it simple and go step by step - transformers annotations for modular and transformers style rope (complex view)
* more cleanup
* refactor namings and remove addition forXXX classes
* our moe won't cut it it seems, correction bias seems to be missing in remote code version
* tokenization change (remote)
* our moe version works when adding normalization :D
* cleanup moe
* nits
* cleanup modeling -> let's get to modular next
* style
* modular v1
* minor things + attempt at conversion (which doesn't work)
* no conversion follow glm, fixup modular and other nits
* modular cleanup
* fixes
* tests, tests, tests + some moe dtype forcing
* simplify modular, fix fatal fa2 bug, remaining tests
* fix import issue?
* some initial docs, fix bnb faulty behavior --> needs to fix some tests because of gate needing to be float
* fix sdpa test, load on init dtype only
* fixup post merge
* style
* fix doc links
* tokenization cleanup beginnings
* simplify tokenizer by a lot as its basically llama
* tokenizer is full llama with different defaults + extra special tokens
* sync og special tokens of ernie
* fix decoding with numbers (also in remote done what a timing), begin of tok tests
* align with remote and preserve special tokens, adjust tests to ernie legacy behavior, warning for questionable behavior (also in llama)
* nits
* docs
* my daily post merge it is
* check
* tokenization update with explanations and conversion script
* review on modular (til), revert some tokenizer things i did prior, remove mtp comment (low prio)
* post merge fixes
* fixup tokenization, llama fast is the way to go
* more fixups
* check
* import fixes
* correction bias following the paddle code
* fix
* fix TP plan, fix correction bias sharding during forward
* style
* whoops
* fix tied weights
* docs and last nit
* license
* flasky tests
* move repo id, update when merged on the hub
* fix qwen2 vl packing in FA2
* why? delete!
* qwen2-5-vl seems to work now
* update
* fix tests
* start by adapting FA2 tests
* add similar tests for sdpa/eager
* address comments
* why is this even in conditional model and not base model?
* dump
* push other models
* fix simple greedy generation
* xmod
* add fmst and clean up some mentions of old cache format
* gpt-bigcode now follows standards
* delete tuple cache reference in generation
* fix some models
* fix some models
* fix mambas and support cache in tapas
* fix some more tests
* fix copies
* delete `_reorder_cache`
* another fix copies
* fix typos and delete unnecessary test
* fix rag generate, needs special cache reordering
* fix tapas and superglue
* reformer create special cache
* recurrent gemma `reorder_cache` was a no-op, delete
* fix-copies
* fix blio and musicgen pipeline tests
* fix reformer
* fix reformer, again...
* delete `_supports_cache_class`
* delete `supports_quantized_cache`
* fix failing tests
* fix copies
* some minor clean up
* style
* style
* fix copies
* fix tests
* fix copies
* create causal mask now needs positions?
* fixc copies
* style
* Update tests/test_modeling_common.py
Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>
* clean-up of non-generative model after merging main
* check `is_decoder` for cache
* delete transpose for scores
* remove tuple cache from docs everywhere
* fix tests
* fix copies
* fix copies once more
* properly deprecate `encoder_attention_mask` in Bert-like models
* import `deprecate_kwarg` where needed
* fix copies again
* fix copies
* delete `nex_decoder_cache`
* fix copies asks to update for PLM
* fix copies
* rebasing had a few new models, fix them and merge asap!
* fix copies once more
* fix slow tests
* fix tests and updare PLM checkpoint
* add read token and revert accidentally removed line
* oh com -on, style
* just skip it, read token has no access to PLM yet
---------
Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>
* working locally; need to style and test
* added docs and initial tests; need to debug and flesh out
* fixed tests
* working long context; batches
* working fa2 and eager
* update tests
* add missing confnigs
* remove default autoset
* fix spacing
* fix most tests
* fixed tests
* fix to init
* refactor to match new transformers updates
* remove static cache option
* fa2 fix
* fix docs
* in progress
* working on tests
* fixed issue with attn outputs
* remove debug
* fix local config attr
* update doc string
* fix docstring
* add docs to toc
* correct typo in toc
* add new updates from main w.r.t. ModernBERT RoPE
* fix local param
---------
Co-authored-by: oweller2 <oweller2@dsailogin.mgmt.ai.cluster>
Co-authored-by: oweller2 <oweller2@l07.mgmt.ai.cluster>
Co-authored-by: oweller2 <oweller2@n02.mgmt.ai.cluster>
Co-authored-by: oweller2 <oweller2@l08.mgmt.ai.cluster>
Co-authored-by: oweller2 <oweller2@l01.mgmt.ai.cluster>
Co-authored-by: oweller2 <oweller2@l02.mgmt.ai.cluster>
* just update 2 files
* update other models as well just making fix-copies
* also add the changes needed to modeling utils
* put this on the pretrained model instead
* nits and fixes
* update generic, fix to use config value
* update other modelings
* use transformers kwargs instead
* update
* update
* update other models
* update
* updates
* update
* update
* update
* fix
* finally
* very small nits
* this fixes more tests
* fix other models as well!
* update modularqwen2
* update models based on qwen2
* update
* update
* remove the **flash stuff in favor of noraml kwargs
* update
* propagate gemma?
* remove output attentions
* propagate
* support cross attention edge case
* same
* test this
* fixes
* more fix
* update
* update
* fix conflicts
* update
* fix emu3
* fix emu3
* move the fix a bit
* quel enfer
* some fixes, loss_kwargs should never had been
* finish fixing gemma3n
* fix small lm3
* fix another one
* fix csm now
* fux csm and mistral
* fix mistral now
* small fixes
* fix janusss
* only for some models
* fixup
* phix phi3
* more fixes?
* dose this fix it?
* update
* holy shit it was just graph breaks
* protect torch
* updates
* fix samhq?
* fix moonshine
* more moonshine fixes, 3 failures left!
* nits
* generic needs to support more
* more fixes to moonshine!
* fix cross attention outputs!
* fix csm!
* nits
* fix stupid kosmos2
* current updates
* fixes
* use output recorder?
* nicer!
* a little bit of magic
* update
* fix protect
* fix
* small fixes
* protect import
* fix a bunch of more models
* fix fixups
* fix some of the last ones
* nit
* partly fix phi
* update
* fix import path
* make something that is fullgraph compatible just to be sure
* typing was wrong on llama so the rest was wrong as well
* fucking ugly but at least it is still exportable
* syle
* supposed to fix moonshine, it still breaks
* fix some default
* fix the last bits of sam
* update samhq
* more fixes to am hq
* nit
* fix all output+hidden states and output_attentions!
* fix?
* fix diffllama
* updates to fix initialization on the sam pips
* ups there was a bug
* fix the last sam hq test
* fix gotocr
* fix gotocr2!
* fixes
* skip stupid tests
* there was one left :)
* fixup
* fix fix copies issues with this test file
* fix copies for sam_hq
* rm some comments
* skip 2 more failing tests
* fix
* fix everything
* Apply suggestions from code review
Co-authored-by: Anton Vlasjuk <73884904+vasqu@users.noreply.github.com>
Co-authored-by: Pablo Montalvo <39954772+molbap@users.noreply.github.com>
* add more doc!
* fix public init
* fix modular qwen3
---------
Co-authored-by: Anton Vlasjuk <73884904+vasqu@users.noreply.github.com>
Co-authored-by: Pablo Montalvo <39954772+molbap@users.noreply.github.com>
* remove the skips
* fix the epsilon to a small value (does not make sense otherwise)
* safeguard
* overload test_eager_matches_sdpa
* Update test_modeling_common.py
* skip appropriate tests
* correct no_split_layer
* fix all devices issue
* fix backward
* fix
* rm tf/flax tests
* more flax deletions
* revert fixture change
* reverted test that should not be deleted; rm tf/flax test
* revert
* fix a few add-model-like tests
* fix add-model-like checkpoint source
* a few more
* test_get_model_files_only_pt fix
* fix test_retrieve_info_for_model_with_xxx
* fix test_retrieve_model_classes
* relative paths are the devil
* add todo
* add dia model
* add tokenizer files
* cleanup some stuff
* brut copy paste code
* rough cleanup of the modeling code
* nuke some stuff
* more nuking
* more cleanups
* updates
* add mulitLayerEmbedding vectorization
* nits
* more modeling simplifications
* updates
* update rope
* update rope
* just fixup
* update configuration files
* more cleanup!
* default config values
* update
* forgotten comma
* another comma!
* update, more cleanups
* just more nits
* more config cleanups
* time for the encoder
* fix
* sa=mall nit
* nits
* n
* refacto a bit
* cleanup
* update cv scipt
* fix last issues
* fix last nits
* styling
* small fixes
* just run 1 generation
* fixes
* nits
* fix conversion
* fix
* more fixes
* full generate
* ouf!
* fixes!
* updates
* fix
* fix cvrt
* fixup
* nits
* delete wrong test
* update
* update
* test tokenization
* let's start changing things bit by bit - fix encoder step
* removing custom generation, moving to GenerationMixin
* add encoder decoder attention masks for generation
* mask changes, correctness checked against ad29837 in dia repo
* refactor a bit already --> next cache
* too important not to push :)
* minimal cleanup + more todos
* make main overwrite modeling utils
* add cfg filter & eos filter
* add eos countdown & delay pattern
* update eos countdown
* add max step eos countdown
* fix tests
* fix some things
* fix generation with testing
* move cfg & eos stuff to logits processor
* make RepetitionPenaltyLogitsProcessor flexible
- can accept 3D scores like (batch_size, channel, vocab)
* fix input_ids concatenation dimension in GenerationMixin for flexibility
* Add DiaHangoverLogitsProcessor and DiaExponentialDecayLengthPenalty classes; refactor logits processing in DiaForConditionalGeneration to utilize new configurations and improve flexibility.
* Add stopping criteria
* refactor
* move delay pattern from processor to modeling like musicgen.
- add docs
- change eos countdown to eos delay pattern
* fix processor & fix tests
* refactor types
* refactor imports
* format code
* fix docstring to pass ci
* add docstring to DiaConfig & add DiaModel to test
* fix docstring
* add docstring
* fix some bugs
* check
* porting / merging results from other branch - IMPORTANT: it very likely breaks generation, the goal is to have a proper forward path first
* experimental testing of left padding for first channel
* whoops
* Fix merge to make generation work
* fix cfg filter
* add position ids
* add todos, break things
* revert changes to generation --> we will force 2d but go 3d on custom stuff
* refactor a lot, change prepare decoder ids to work with left padding (needs testing), add todos
* some first fixes to get to 10. in generation
* some more generation fixes / adjustment
* style + rope fixes
* move cfg out, simplify a few things, more todos
* nit
* start working on custom logit processors
* nit
* quick fixes
* cfg top k
* more refactor of logits processing, needs a decision if gen config gets the new attributes or if we move it to config or similar
* lets keep changes to core code minimal, only eos scaling is questionable atm
* simpler eos delay logits processor
* that was for debugging :D
* proof of concept rope
* small fix on device mismatch
* cfg fixes + delay logits max len
* transformers rope
* modular dia
* more cleanup
* keep modeling consistently 3D, generate handles 2D internally
* decoder starts with bos if nothing
* post processing prototype
* style
* lol
* force sample / greedy + fixes on padding
* style
* fixup tokenization
* nits
* revert
* start working on dia tests
* fix a lot of tests
* more test fixes
* nit
* more test fixes + some features to simplify code more
* more cleanup
* forgot that one
* autodocs
* small consistency fixes
* fix regression
* small fixes
* dia feature extraction
* docs
* wip processor
* fix processor order
* processing goes brrr
* transpose before
* small fix
* fix major bug but needs now a closer look into the custom processors esp cfg
* small thing on logits
* nits
* simplify indices and shifts
* add simpler version of padding tests back (temporarily)
* add logit processor tests
* starting tests on processor
* fix mask application during generation
* some fixes on the weights conversion
* style + fixup logits order
* simplify conversion
* nit
* remove padding tests
* nits on modeling
* hmm
* fix tests
* trigger
* probably gonna be reverted, just a quick design around audio tokenizer
* fixup typing
* post merge + more typing
* initial design for audio tokenizer
* more design changes
* nit
* more processor tests and style related things
* add to init
* protect import
* not sure why tbh
* add another protect
* more fixes
* wow
* it aint stopping :D
* another missed type issue
* ...
* change design around audio tokenizer to prioritize init and go for auto - in regards to the review
* change to new causal mask function + docstrings
* change ternary
* docs
* remove todo, i dont think its essential tbh
* remove pipeline as current pipelines do not fit in the current scheme, same as csm
* closer to wrapping up the processor
* text to audio, just for demo purposes (will likely be reverted)
* check if it's this
* save audio function
* ensure no grad
* fixes on prefixed audio, hop length is used via preprocess dac, device fixes
* integration tests (tested locally on a100) + some processor utils / fixes
* style
* nits
* another round of smaller things
* docs + some fixes (generate one might be big)
* msytery solved
* small fix on conversion
* add abstract audio tokenizer, change init check to abstract class
* nits
* update docs + fix some processing :D
* change inheritance scheme for audio tokenizer
* delete dead / unnecessary code in copied generate loop
* last nits on new pipeline behavior (+ todo on tests) + style
* trigger
---------
Co-authored-by: Arthur Zucker <arthur.zucker@gmail.com>
Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
Co-authored-by: Vasqu <antonprogamer@gmail.com>
* Support `flash_attn_3`
Implements fwd and tests for Flash Attention 3 https://github.com/Dao-AILab/flash-attention/commits/main/hopper
- Includes checks for dropout>0 and ALiBi in `modeling_utils.PreTrainedModel._check_and_enable_flash_attn_3` (Dropout will likely be supported soon, so this will need to be updated and `modeling_flash_attention_utils._flash_attention_forward` at the `if _IS_FLASH_ATTN_3_AVAILABLE: ...`
An example Llama implementation is included in `modeling_llama.py` but other models would still need to be updated
Based on https://github.com/huggingface/transformers/pull/36190 which has model implementations and examples which could be merged
* Add tests for Flash Attention 2 and 3 parity
* ci fix
* FA2 compatibiity
- `_prepare_flash_attention_from_position_ids` ->`prepare_fa2_from_position_ids`
- Remove bettertransformer check in Flash Attention 3
- Merge tests
- Add licensing
* ci fix
* Test naming consistency
* ci fix
* Deprecation warning for `prepare_fa2_from_position_ids`
* ci fix
* Fix graph break in torch.compile when using FA2 with attention_mask=None and batch size > 1
* fix code format
* add test; replace position_ids with query_states becasue position_ids.shape[0] is always 1
* add assert loss is not nan
* remove it from all py files
* remove it from the doc
* remove it from examples
* style
* remove traces of _fast_init
* Update test_peft_integration.py
* CIs
* adding model and conversion scripts
* add imports to test vjepa conversion
* fix imports and make conversion work
* fix computation for short side
* replace attention with library attention function
* cleanup more attention classes
* remove config overrides
* add test cases, fix some of the failing ones
* fix the model outputs
* fix outputs of the model per review
* fix too big model test case
* fix styling __init__.py
* fix initialization test
* remove all asserts per review
* update sorting unsorting logic as per feedback
* remove is_video per review
* remove another is_video segment
* remove unwanted stuff
* small fixes
* add docstrings for the model
* revert adding vjepa2 config here
* update styling
* add config docstrings (wip)
* fix dpr issue
* removed test failing issues
* update styles
* merge predictor configs into main config
* remove processing code, add video processor
* remove permute which is not necessary now
* fix styles
* updated vjepa2 to be in video_processing_auto
* update comment for preprocessing
* test integration test and fix the outputs
* update test values, change test to look at repeated frames for a given image
* add a simple video processing test
* refactoring pixel_values_videos and upload ckpts to original
* fix torch_fx test cases
* remove unused config
* add all config docstrings
* add more integration tests
* add basic doc
* revert unwanted styling changes
* working make fixup
* Fix model_type in config
* update attention implementation to fit new hf standards
* fix the preprocessing logic, ensure it matches the original model
* remove use_rope logic, cleanup
* fix docstrings
* Further cleanup, update doc
* Fix model prefix
* fix get_vision_features
* VJEPA2Embeddings style refactor
* nit, style comment
* change modules default values
* Only `str` activation in config
* GradientCheckpointingLayer
* fixup
* fix conversion script
* Remove return_dict
* remove None return typehint
* Refactor VJEPA2Layer, remove use_SiLU
* Fix fx tests
* dpr -> drop_path_rates
* move *ModelOutput on top
* format docs bit
* update docs
* update docs
* update doc example
* remove prune_heads from model
* remove unused config params
* refactor embed signature
* Add vjepa to docs
* Fix config docstring
* update defaults
* Update docs/source/en/model_doc/vjepa2.md
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>
* Update docs/source/en/model_doc/vjepa2.md
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>
* Fix import
* Min refactoring
* Update HUB_SOURCE and HUB_REPO in conversion script
* Add missing headers
* VJEPA -> V-JEPA in docs
* Add image to doc
* fix style
* fix init weights
* change checkpoint name in modeling tests
---------
Co-authored-by: Koustuv Sinha <koustuv.sinha@mail.mcgill.ca>
Co-authored-by: yonigozlan <yoni.gozlan@huggingface.co>
Co-authored-by: Yoni Gozlan <74535834+yonigozlan@users.noreply.github.com>
Co-authored-by: Koustuv Sinha <koustuvsinha@gmail.com>
Co-authored-by: Pedro Cuenca <pedro@huggingface.co>
* start refactoring whisper
* revert for now
* first step
* carry over attn fixes
* check if this works
* whisper has an off by one somewhere - cutting mask in any interface
* make it based on interface
* remove some tests that were skipped but now work
* some fixes for whisper tests
* interface changes
* change the order of fix
* some attention adjustments for eager + TP
* fix scaling
* mask changes
* why does whisper contain those extra seq lens?
* fix from config for fa2 as input_ids is invalid
* fix another test
* another fix
* disable flex attn due to compile issues
* copies and refactor for qwen audio since it somewhat relies on whisper
* fix scaling and smaller things
* retrigger
* new new interface version + more fixups
* adjust qwen
* add comment
* forgot this one
* change copies as whisper cuts on the mask
* add guard
* add flex attention
* switch to new mask function + add skips for torchscript
* remove old api with cache position
* last changes?
* trigger ci
