diff --git a/Dockerfile b/Dockerfile index c9393bf..44e0a21 100644 --- a/Dockerfile +++ b/Dockerfile @@ -43,4 +43,7 @@ RUN apt-get update && apt-get install -y git \ #COPY hf.py /usr/local/lib/python3.12/dist-packages/vllm/renderers/hf.py # Make sure we have the latest up to date chat template -#COPY glm_5.1_chat_template.jinja /opt/chat_template.jinja \ No newline at end of file +COPY glm_5.1_chat_template.jinja /opt/chat_template.jinja + +# Working LMCache config +COPY lmcache-config.yaml /opt/lmcache-config.yaml \ No newline at end of file diff --git a/lmcache-config.yaml b/lmcache-config.yaml new file mode 100644 index 0000000..1229f3c --- /dev/null +++ b/lmcache-config.yaml @@ -0,0 +1,7 @@ +chunk_size: 256 +local_cpu: true +max_local_cpu_size: 256.0 +save_decode_cache: false +enable_lazy_memory_allocator: true +lazy_memory_initial_ratio: 1.0 +use_gpu_connector_v3: true