Do you have a plan to implement the CudaCatOp?

ztxz16 / fastllm

纯c++的全平台llm加速库，支持python调用，chatglm-6B级模型单卡可达10000+token / s，支持glm, llama, moss基座，手机端流畅运行

Apache License 2.0

3.33k stars 341 forks source link

Open dp-aixball opened 7 months ago