Release v0.2.4
Added
- Tuned the hyperparameters of DiodeMix optimizer for sft.
- Added sft-support for the classical gptq-style models.
- Implemented qzeros update in finetuning process.
Updated
- Extended pack_fp_weight function.
- Enhanced the performance of MPQLinearCUDA layer.
Fixed
- Fixed various errors in DiodeMix update function.