feature: add

docker-images/Dockerfile

@@ -1,8 +0,0 @@
-FROM harbor.4pd.io/mic-llm-x/combricon_vllm_mlu:v1.0_0510
-
-WORKDIR /workspace
-
-COPY launch_service /workspace
-
-
-CMD ["bash", "launch_service"]

docker-images/a100.dockerfile

@@ -0,0 +1,8 @@
+FROM harbor.4pd.io/sagegpt-aio/pk_platform/vllm-installed:v0.10.0
+
+WORKDIR /workspace
+
+COPY a100_launch_service /workspace
+COPY template.jinja /workspace
+
+ENTRYPOINT ["bash","/workspace/a100_launch_service"]

docker-images/a100_launch_service

@@ -7,7 +7,7 @@ export
 date
 
 DEFAULT_HOST="0.0.0.0"
-DEFAULT_PORT="8000"
+DEFAULT_PORT="80"
 DEFAULT_SERVED_MODEL_NAME="llm"
 DEFAULT_MODEL_PATH="/model"
 DEFAULT_MAX_MODEL_LEN="8192"
@@ -47,7 +47,7 @@ CMD_ARGS+=(--model "$MODEL_PATH_VAL")
 CMD_ARGS+=(--max-model-len "$MAX_MODEL_LEN_VAL")
 CMD_ARGS+=(--tensor-parallel-size "$TENSOR_PARALLEL_SIZE_VAL")
 CMD_ARGS+=(--max-num-seqs "$MAX_NUM_SEQS_VAL")
-CMD_ARGS+=(--trust-remote-code --dtype float16 --block-size 8192)
+CMD_ARGS+=(--trust-remote-code --dtype float16 --chat-template /workspace/template.jinja)
 echo "--------------------------------------------------"
 echo "Starting VLLM OpenAI API Server..."
 echo "Using effective arguments:"
@@ -66,4 +66,3 @@ echo "python3 -m vllm.entrypoints.openai.api_server ${CMD_ARGS[*]}"
 echo "--------------------------------------------------"
 
 python3 -m vllm.entrypoints.openai.api_server "${CMD_ARGS[@]}"
-

docker-images/mlu370.dockerfile

@@ -0,0 +1,7 @@
+FROM harbor.4pd.io/mic-llm-x/combricon_vllm_mlu:v1.0_0510
+
+WORKDIR /workspace
+
+COPY mlu370_launch_service /workspace
+
+ENTRYPOINT ["bash","/workspace/mlu370_launch_service"]