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
dynamic map checkpoint naming based on model objective. (#1840)
6c76487 Xin He committed 5 months ago
Security: HTTP requests are performed without timeout safeguards (#1683)
fbff417 tomaioo committed 5 months ago
Reduce RAM usage of quantizing VLM models and fix some issues of quantizing gemma4 (#1791)
9696884 Liang Lv committed 6 months ago
Update dependencies in CI and installation scripts for cuda compatibility (#1825)
bd0a1f1 Sun, Xuehao committed 6 months ago
[Fix Accuracy] fix layer name mismatch of VLM(qwen3.5-2B) in hf loading (#1823)
425a6ff Xin He committed 6 months ago
Feats: Quantize/save/evaluate the Wan-AI/WAN2.2 models in w4a16 format (#1678)
f1db02a Liang Lv committed 6 months ago
[bug fix] model free: fix copy folder of diffusion model, fix unresponsive disable_model_free, and cuda UT (#1809)
2c915bb Xin He committed 6 months ago
chore: clean wheel download directory and update path exclusions in test configurations (#1822)
e46231b chen, suyue committed 6 months ago
Support ByteDance-Seed/BAGEL-7B-MoT quantization in w4a16 format (#1633)
91679a5 Liang Lv committed 6 months ago
Scope ARK oneDNN XPU build, bump ARK version to 0.13.1, and remove the dpcpp runtime dependency (#1819)
a9f8fc1 Copilot committed 6 months ago