GitLab

Feat absorb for long prefill

[feat] Support fp8 linear kernel;

musa: support bf16

Ensure backward compatibility with PyTorch 2.2

Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>

fix KExpertsMarlin on GPU with out CUDA Graph

support moonlight, use ktransformers/optimize/optimize_rules/Moonlight-16B-A3B.yaml

fix bf16 load, TODO: refactor cpu dequant

Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>

Add fp8 linear kernel;\n Add empty cache to fit in 16G VRAM; By 'wkGCaSS - 知乎 https://zhuanlan.zhihu.com/p/25491611225'

optimize CMake multi core parallel

Feat more context

Fix the link address in the doc install.md

Adjust the installation link to the correct section of docs

use marlin for lm_head, lm_head only calc last token for prefill
extend context window to 19K for DeepSeek-V3/R1 within 24GB VRAM

feat: Support Moore Threads GPU

Docker dev

feat: add GitHub Actions workflow for building Docker image