GitLab

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

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

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

1) Linear and MLP operators support qlen>1; 2) All operators now share a single memory buffer; 3) Refactor CPUInfer submit/sync logic.

support windows support q4_0 and q5_0 dequant on cpu Add CopyRight from pygguf(It was added before, but disappear after merge). Add some TODO in the code.