transformers

transformers copied to clipboard

fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP is not working with the Trainer

5

### System Info acc_cfg.yml: compute_environment: LOCAL_MACHINE debug: false distributed_type: FSDP downcast_bf16: 'no' enable_cpu_affinity: true fsdp_config: fsdp_activation_checkpointing: true fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP fsdp_backward_prefetch: NO_PREFETCH fsdp_cpu_ram_efficient_loading: true fsdp_forward_prefetch: true fsdp_offload_params: true fsdp_sharding_strategy: FULL_SHARD fsdp_state_dict_type:...

eljandoubi

Finetuned LLAMA Model is working same old pretrained model after combining LORA weights with old model

2

I have finetuned a LLAMA-7b-chat-hf model and saved the adaptor weights. After Loading and merging the adaptor weights to old model, both new model and old model was giving me...

abhi201002

Add propainter

34

# What does this PR do? This PR adds ProPainter, a Video Inpainting model with 5.4k stars and 635 forks [repo](https://github.com/sczhou/ProPainter). It fixes #26360 and resolve stale PR #26391 for...

RUFFY-369

Add GGUF for Mamba

1

# What does this PR do? Add GGUF support for Mamba ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the...

VladOS95-cyber

# What does this PR do? Add [ColPali](https://doi.org/10.48550/arXiv.2407.01449) support in 🤗 `transformers`. ## Who can review? @yonigozlan 😉 ## Additional details - This PR uses the new [Modular 🤗 transformers](https://huggingface.co/docs/transformers/main/en/modular_transformers#modular-transformers)...

tonywu71

Add Segment Anything 2 (SAM2)

9

# What does this PR do? https://github.com/huggingface/transformers/issues/32308 As stated in this issue this PR is making SAM2 compatible to transformers cc. @zinccat @RUFFY-369 Fixes # (issue) ## Before submitting -...

SangbumChoi

# What does this PR do? In [#34198](https://github.com/huggingface/transformers/commit/6ba31a8a94bf7cfeaf59ffc3bc9e0b0cd3e25788#diff-ed55888e6665791fe92cc8fc0c499da54f4ace6738551cd9a2591881cda076deR3629), the line `loss *= self.args.gradient_accumulation_steps` was introduced due to `Negate accelerate grad accum div`. This change was made to correct errors encountered...

BlackNoodle

New GA fix causes training loss multiple times higher across the board (5x to 10x higher)

1

### System Info 8xH100 ### Who can help? _No response_ ### Information - [ ] The official example scripts - [ ] My own modified scripts ### Tasks - [...

JianbangZ

Request for a clear documentation for .generate()

2

### Feature request The `.generate()` function has a lot of parameters, for example `length_penalty` and `diversity_penalty`. However, the [documentation](https://huggingface.co/docs/transformers/v4.45.1/en/main_classes/text_generation#transformers.GenerationMixin.generate) of this function does not document a full list of parameters,...

sedol1339

fix error in _get_eval_sampler when group_by_length enabled

2

# What does this PR do? https://github.com/huggingface/transformers/pull/33514 this PR adds group_by_length support for evaluation. But, this part uses self.eval_dataset instead of eval_dataset. So, if eval_dataset is dictionary it fails. ##...

akakakakakaa