OpenGVLab / OpenGVLab/Ask-Anything
ValueError: Attempting to unscale FP16 gradients when fully finetunning on LLM without LoRA?
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 3.4k
- Forks
- 268
- PR merge metrics
- No merged PRs in 30d
Description
dear all:
when i disable the lora and set whole Mistral LLm weights requires_grad=True, it gives me the error:
ValueError: Attempting to unscale FP16 gradients
anyone give some tips?
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by reproducing the reported error with the Mistral LLM fully fine-tuned and LoRA disabled, then trace the FP16 gradient unscaling path. Confirm the training configuration and determine what change allows full fine-tuning to proceed without the ValueError.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100