LoRA PEFT #50
Merged
LoRA PEFT #50
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
commit dd8de48 Author: Neil Tan <[email protected]> Date: Mon Mar 24 11:47:51 2025 +0800 Modified README.md moved lora peft code to gr00t/utils/peft.py added peft dependency added lora_rank argument to scripts/gr00t_petf_finetune.py commit 569ff64 Author: Neil Tan <[email protected]> Date: Mon Mar 24 05:08:02 2025 +0800 commandline arg r=32 commit 4c8067b Author: Neil Tan <[email protected]> Date: Sun Mar 23 13:39:04 2025 +0800 only wrapping forward() commit 4fa6051 Author: Neil Tan <[email protected]> Date: Sun Mar 23 13:08:34 2025 +0800 seems to run
youliangtan
requested changes
Mar 24, 2025
There was a problem hiding this comment.
Thanks for the PR! This is very helpful for efficient finetuning on comsumer grade GPU.
First could you run a styling with black .?
Also i noticed this error when running inference after merging #39 which set_pretrained(..., torch_dtype='auto')
Other than that it works well!
Model not found or avail in the huggingface hub. Loading from local path: /tmp/gr00t/checkpoint-20/
Traceback (most recent call last):
File "/home/youliang/gear/clean_repo/Isaac-GR00T/scripts/inference_service.py", line 73, in <module>
policy = Gr00tPolicy(
File "/home/youliang/gear/clean_repo/Isaac-GR00T/gr00t/model/policy.py", line 106, in __init__
self._load_model(model_path)
File "/home/youliang/gear/clean_repo/Isaac-GR00T/gr00t/model/policy.py", line 232, in _load_model
model = GR00T_N1.from_pretrained(model_path, torch_dtype="auto")
File "/home/youliang/gear/clean_repo/Isaac-GR00T/gr00t/model/gr00t_n1.py", line 224, in from_pretrained
pretrained_model = super().from_pretrained(
File "/home/youliang/miniconda3/envs/groot-release/lib/python3.10/site-packages/transformers/modeling_utils.py", line 3848, in from_pretrained
dtype_orig = cls._set_default_torch_dtype(torch_dtype)
File "/home/youliang/miniconda3/envs/groot-release/lib/python3.10/site-packages/transformers/modeling_utils.py", line 1617, in _set_default_torch_dtype
if not dtype.is_floating_point:
AttributeError: 'str' object has no attribute 'is_floating_point'
|
@youliangtan |
youliangtan
previously requested changes
Mar 24, 2025
- removed verbose lora module print outs - updated lora fine-tune hyperparameters in the README.md
|
All good now. Thanks for the PR |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR enable LoRA PEFT to be enabled via an additional command line argument for the fine-tuning example.
Changes proposed in this pull request:
- Updated README.md
- moved lora peft code to gr00t/utils/peft.py
- added peft installation dependency
- added lora_rank argument to scripts/gr00t_petf_finetune.py