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
[Experimental]Add MLX format export support and AutoScheme for vlm support (#1732)
0def9d6 Wenhua Cheng committed 6 months ago
Reduce xpu memory usage with patch_xpu_sdpa_drop_causal_mask (#1716)
dc8a6a4 Xin He committed 6 months ago
Revert "ignore mtp.fc for qwen3_5 due to vllm failure (#1710)" (#1730)
c234455 Xin He committed 7 months ago
[research feature] Introduce INT4 support at the algorithm level (#1641)
11e9186 Wenhua Cheng committed 7 months ago
Support diffusion model AIDC-AI/Ovis-Image-7B quantization (#1616)
95a84d2 Liang Lv committed 7 months ago
Fix hadamard transform weight dtype, using float32 as default and in-place transformed weight. (#1665)
43748dd lkk committed 7 months ago
Stabilize Qwen3-Omni MoE weight fidelity test for matching NaNs (#1681)
865f31b Liang Lv committed 7 months ago
Enable NextStepDiffusion and support multi-device tuning for diffusion (#1640)
f81e20d Xin He committed 7 months ago