LLaVA INT4 quantization
Open
- Dominant language
- Python
- Stars
- 8.1k
- Forks
- 748
- Avg merge
- 6d 2h
- Merged PRs (30d)
- 54
Description
The performance of the LLaVA-1.5 13B model drops significantly after using 4-bit quantization. After turning on –search-scale, the performance still drops significantly.
Contributor guide
Assessment
This issue has not been assessed yet.