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
Automatically select calibration datasets for code models (#2107)
f8de00c Wang, Chang committed 3 months ago
Fix: temporarily skip tests due to known issues with xfail annotations (#2113)
60b813c chen, suyue committed 3 months ago
[Feature] Register_block_output to support GLM5.2 tuning (#2104)
d05fb63 Xin He committed 3 months ago
Enable torch.compile by default and update related documentation (#2070)
be862d3 Xin He committed 3 months ago
Stream AutoScheme's per-layer sensitivity scoring block-by-block (#2063)
73661ed Fabrizio del Tin committed 3 months ago
Fix dataset and batch_size not taking effect in CLI (#2102)
212c8bd Wang, Chang committed 3 months ago
[Feature]: speedup autoscheme with multi-process and cache files (#2083)
288f848 Xin He committed 3 months ago
Fix Flux diffusion state propagation during block replay (#2086)
3cae4b6 Wang, Chang committed 3 months ago
[high-risk][Step 1]Refactor compressor and quantizer (#2039)
8f8c8e9 Wenhua Cheng committed 4 months ago
fix: AutoScheme embedding budget and gguf mixed-bits config for multimodal models (#2055)
06c33ea Heng Guo committed 4 months ago
fix: add triton to requirements for LLMC CPU tests (#2054)
c774c83 Sun, Xuehao committed 4 months ago
enhance logging and data loading for AutoScheme and model-free compatibility (#2050)
6f45d22 Xin He committed 4 months ago
fix: handle router_logits tuple for compatibility with transformers >= 5.13.0 (#2046)
4b8468a Xin He committed 4 months ago
Fix invalid group reference in revert_checkpoint_conversion_mapping (#2035)
6d1c6db A Sivasubramanian Manoj committed 4 months ago