Skip to content

Conversation

zhuohan123
Copy link
Member

@zhuohan123 zhuohan123 commented Apr 30, 2023

Fix #48.

This PR makes the numpy copy in the previous weight loading optional. Specifically, we implement a new hf_model_weights_iterator, which iterates all the weights of a huggingface checkpoint. We then load each weight in the checkpoint to the model's state_dict.

@zhuohan123 zhuohan123 requested a review from WoosukKwon May 2, 2023 09:57
Copy link
Collaborator

@WoosukKwon WoosukKwon left a comment

Choose a reason for hiding this comment

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

Awesome. Left a comment about minor refactoring.

@zhuohan123 zhuohan123 merged commit 27f1410 into main May 3, 2023
@zhuohan123 zhuohan123 deleted the new-weight-loader branch May 24, 2023 04:40
hongxiayang pushed a commit to hongxiayang/vllm that referenced this pull request Feb 13, 2024
dllehr-amd pushed a commit to dllehr-amd/vllm that referenced this pull request Jul 22, 2024
* fix __init__ files

* make yapf happy
JHLEE17 pushed a commit to JHLEE17/vllm that referenced this pull request Aug 1, 2024
@alixiaodi alixiaodi mentioned this pull request Aug 2, 2024
wuhuikx pushed a commit to wuhuikx/vllm that referenced this pull request Mar 27, 2025
### What this PR does / why we need it?
fix communicator patch so parallel could work.
see vllm-project#52

Signed-off-by: MengqingCao <[email protected]>
heheda12345 pushed a commit to heheda12345/vllm that referenced this pull request Sep 29, 2025
Bounty-hunter added a commit to Bounty-hunter/vllm that referenced this pull request Sep 30, 2025
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.

Improve Weight Loading

2 participants