* Add check for target_sizes is None in post_process_image_guided_detection
* Make sure Owlvit and Owlv2 in sync
* Fix incorrect indentation; add check for correct size of target_sizes
* don't log base model architecture in wandb is log model is false
* Update src/transformers/integrations/integration_utils.py
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>
* convert log model setting into an enum
* fix formatting
---------
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>
* use currently active microphone on mac for ffmpeg_microphone
* Allow ffmpeg_microphone device to be specified
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>
---------
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>
* No more default chat templates
* Add the template to the GPT-SW3 tests since it's not available by default now
* Fix GPT2 test
* Fix Bloom test
* Fix Bloom test
* Remove default templates again
* fix: default value reflects the runtime environment variables rather than the ones present at import time.
* Fix: Change `deterministic` to None by default; use env var if None
* Updated ruff version and fixed the required code accorindg to the latest version.
* Updated ruff version and fixed the required code accorindg to the latest version.
* Added noqa directive to ignore 1 error shown by ruff
* add DataCollatorBatchFlattening
* Update data_collator.py
* change name
* new FA2 flow if position_ids is provided
* add comments
* minor fix
* minor fix data collator
* add test cases for models
* add test case for data collator
* remove extra code
* formating for ruff check and check_repo.py
* ruff format
ruff format tests src utils
* custom_init_isort.py
* feat(cache): StaticCache uses index_copy_ to avoid useless copy
Using index_copy_ allows for explicit in-place change of the tensor.
Some backends (XLA) will otherwise copy the tensor, making the code
slower and using more memory.
Proposed implementation will end up using less memory and on XLA will
result in less compilation, but the change is also quite generic, making
no change whatsoever on CUDA or CPU backend.
* feat(cache): SlidingWindowCache uses index_copy_ to avoid useless copy
Applying the same change done in StaticCache.
* fix(cache): fallback of index_copy_ when not implemented
* fix(cache): in index_copy_ ensure tensors are on same device
* [run slow] llama
* fix(cache): add move of cache_position to same device in SlidingWindowCache
* Revert "[run slow] llama"
This reverts commit 02608dd142.
* gguf conversion forces add_prefix_space=False for llama3, this is not required and forces from_slow, which fails. changing to None + test
* typo
* clean test
* Change resize_token_embeddings to make it return same Class that is passed to it
* Add explanatory comment as requested in review
* Add explanatory comments for add resizing function in lxmert
* Add comment for padding_idx and moving _resize_bias in lxmert to LxmertForPreTraining
---------
Co-authored-by: Prashanth Sateesh <prasatee@Prashanths-MBP.attlocal.net>
Co-authored-by: Prashanth Sateesh <prasatee@Prashanths-MacBook-Pro.local>
* Add YaRN and Dynamic-YaRN RoPE Scaling Methods
YaRN (Yet another RoPE extension method) combines the NTK-By-Parts
Interpolation and Attention Scaling methods, improving upon existing
RoPE interpolation methods for longer context window sizes.
Fine-tuned models maintain their original performance across benchmarks
while enabling efficient extrapolation and transfer learning for
quicker convergence, especially in compute-limited environments.
We implement YaRN and Dynamic-YaRN for the following list of models:
- LLaMA
- Falcon
- GPT-NeoX
- Olmo
- Persimmon
- Phi
- StableLM
- OpenLLaMA
New unit tests are added to assert YaRN's correct behavior on both
short and long sequence inputs.
For more details, please refer to https://arxiv.org/abs/2309.00071.
Co-authored-by: Miguel Almeida <miguel.pessanha.almeida@tecnico.ulisboa.pt>
* Refactor YaRN implementation for LLaMA
Iterate on YaRN implementation for LLaMA and remove diff from remaining
models for increased PR modularity.
This commit includes the following changes:
- Merge 'yarn_rope_scaling' and 'rope_scaling' dictionaries
- Remove unnecessary attributes ('extrapolation_factor' and 'finetuned')
from YaRN classes
- Inherit 'forward' method in YaRN classes from superclass
- Rename 'yarn' method to 'compute_yarn_scaling'
- Extend YaRN tests with further assertions
- Fix style inconsistencies
Co-authored-by: Miguel Monte e Freitas <miguelmontefreitas@tecnico.ulisboa.pt>
* Refactor Tensor Building Logic for YaRN
- Comply with the the tensor building logic introduced in #30743
- Add referencing to the optimized Attention Factor equation
- Remove Dynamic YaRN for a more agile deployment
Co-authored-by: mig-mfreitas <mig-mfreitas@users.noreply.github.com>
* remove unwanted file
---------
Co-authored-by: Miguel Almeida <miguel.pessanha.almeida@tecnico.ulisboa.pt>
Co-authored-by: mig-mfreitas <mig-mfreitas@users.noreply.github.com>
Co-authored-by: Joao Gante <joao@huggingface.co>
* fix mask creation of gpt2 and gpt_neox caused by me
* forgot the reshape of masks when shape > 2
* add tests for gpt neox and gpt2
* nit on a comment
