Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

vllm/llama-vision llava calibration infinite loop fix #343

Merged
merged 15 commits into from
Nov 26, 2024
Merged

Conversation

WeiweiZhang1
Copy link
Collaborator

No description provided.

Signed-off-by: Zhang, Weiwei1 <[email protected]>
@wenhuach21 wenhuach21 changed the title calib check func bugfix, refine get_dataloader to support INC3.2 vllm/llama-vision llava calibration bug fix, refine get_dataloader to support INC3.2 Nov 25, 2024
@wenhuach21 wenhuach21 changed the title vllm/llama-vision llava calibration bug fix, refine get_dataloader to support INC3.2 vllm/llama-vision llava calibration bug fix Nov 25, 2024
@wenhuach21 wenhuach21 changed the title vllm/llama-vision llava calibration bug fix vllm/llama-vision llava calibration infinite loop fix Nov 25, 2024
@wenhuach21
Copy link
Contributor

remember to remind user to include this fix in the recipe of llama-vision

Signed-off-by: Zhang, Weiwei1 <[email protected]>
Signed-off-by: Zhang, Weiwei1 <[email protected]>
@wenhuach21 wenhuach21 changed the title vllm/llama-vision llava calibration infinite loop fix vllm/llama-vision llava calibration infinite loop fix stage 1 Nov 26, 2024
Signed-off-by: n1ck-guo <[email protected]>
Signed-off-by: n1ck-guo <[email protected]>
@wenhuach21 wenhuach21 changed the title vllm/llama-vision llava calibration infinite loop fix stage 1 vllm/llama-vision llava calibration infinite loop fix Nov 26, 2024
n1ck-guo and others added 3 commits November 25, 2024 22:17
Signed-off-by: n1ck-guo <[email protected]>
Signed-off-by: n1ck-guo <[email protected]>
Signed-off-by: n1ck-guo <[email protected]>
Signed-off-by: n1ck-guo <[email protected]>
Signed-off-by: n1ck-guo <[email protected]>
if quant_nontext_module and batch_size != 1:
logger.warning(f"batch_size({batch_size}) cannot be used for calibrating non-text modules,"
"reset to 1")
batch_size = 1
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

the gradient_accumualte should be changed too

Signed-off-by: n1ck-guo <[email protected]>
@wenhuach21 wenhuach21 self-requested a review November 26, 2024 09:24
@wenhuach21 wenhuach21 merged commit a7fcb51 into main Nov 26, 2024
13 checks passed
@wenhuach21 wenhuach21 deleted the calib_bugfix branch November 26, 2024 09:24
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants