PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度
Commits on main
Add compressed-tensors format export support for W4A16 and W8A16 (#1669)
bbda38a Huang, Tai committed 7 months ago
Refactor module access to use PyTorch get_submodule / set_submodule (#1590)
58221b4 Adithyan committed 7 months ago
Fix missing extra_config export for unsupported ignore_layers like `mlp.gate` (#1660)
d2a62b8 Liang Lv committed 7 months ago
import gptqmodel in confest to avoid error in transformers (#1668)
106a322 Xin He committed 7 months ago
add gptqmodel 6.0 compatible change and MTP quantization skip warning. (#1663)
ac7aa28 Xin He committed 7 months ago
fix autoscheme accuracy drop bug w/o low_gpu, add CI test (#1658)
c9a0d77 Weiwei committed 7 months ago
Enhance quantization configuration support for mixed precision and schemes in utils and tests (#1643)
4632b2e Xin He committed 7 months ago
[Feature] Enhance dataset preprocessing memory management and fix hash failure (#1621)
7e77043 Xin He committed 7 months ago
skip overwriting the CUDA_VISIBLE_DEVICES and support transformers 5.4 (#1626)
47626ec Xin He committed 7 months ago
Allow disabling copy_missing_tensors_from_source via env var, default enabled (#1623)
1897e5e Xin He committed 7 months ago
Fix clean offload clearing of non-checkpoint tensors (#1583)
b4d34c5 Liang Lv committed 7 months ago