[Feature] how to apply a custom calibration data set during awq quant?
Open
- Dominant language
- Python
- Stars
- 8.1k
- Forks
- 748
- Avg merge
- 6d 2h
- Merged PRs (30d)
- 54
Description
### Motivation
the model i want to quant is internvl2. however, i only know how to use official calibration, such as ptb_text_only. This dataset has a different distribution compared with my sft dataset，especially it is text only. So, i hope to quant with my own multimodal dataset to avoid performance drop.
### Related resources
_No response_
### Additional context
_No response_
Contributor guide
Assessment
This issue has not been assessed yet.