GitLab

[Fix] Fix readme structure.

Adapt Windows

Fix: Wrong type of token list returned by prefill_and_generate

Use cond var to avoid busy loop

typo fix: KMisrtal -> KMistral

[fix] Fix some gpu dequant function doesn't support multi gpu bug

Support IQ4_XS dequantize

[fix] Fix qlen > chunk_size mask is none error

[fix] fix bugs about Qwen2-57B, install requirement, DockerFile

docs: update long_context_introduction.md

accuary -> accuracy

[Fix] Fix problem that ktransformers cannot offload whole layer in cpu

fix(docs): fix broken link

[fix]  Fix readme datas

[feature] release 0.1.3

Update README.md

Set the reminder to set CUDA_HOME and CUDA_PATH in the README to the "Quick Start" section under "Install CUDA".

Fix: None for load config

[fix] f16 dequantize device ignored

[fix] Fix bugs about static cache and server param;