vllm-project / vllm-project/aibrix
StreamLoader use `chunk by chunk` instead of `tensor by tensor`
Open
area/acceleration
area/model-loader
kind/enhancement
priority/important-longterm
- Dominant language
- Go
- Stars
- 5.1k
- Forks
- 694
- Avg merge
- 1d 20h
- Merged PRs (30d)
- 98
Description
### 🚀 Feature Description and Motivation
StreamLoader use `chunk by chunk` instead of `tensor by tensor`. The `chunk by chunk` approach can ensure that the size of each pull thread is equal.
### Use Case
_No response_
### Proposed Solution
_No response_
Contributor guide
Assessment
This issue has not been assessed yet.