align vllm hpu version to latest vllm-fork

.github/workflows/docker/compose/llms-compose-cd.yaml

@@ -23,10 +23,6 @@ services:
     build:
       dockerfile: comps/llms/text-generation/vllm/llama_index/Dockerfile
     image: ${REGISTRY:-opea}/llm-vllm-llamaindex:${TAG:-latest}
-  llm-vllm-llamaindex-hpu:
-    build:
-      dockerfile: comps/llms/text-generation/vllm/llama_index/dependency/Dockerfile.intel_hpu
-    image: ${REGISTRY:-opea}/llm-vllm-llamaindex-hpu:${TAG:-latest}
   llm-predictionguard:
     build:
       dockerfile: comps/llms/text-generation/predictionguard/Dockerfile

.github/workflows/docker/compose/llms-compose.yaml

@@ -24,10 +24,6 @@ services:
     build:
       dockerfile: comps/llms/text-generation/vllm/langchain/Dockerfile
     image: ${REGISTRY:-opea}/llm-vllm:${TAG:-latest}
-  llm-vllm-hpu:
-    build:
-      dockerfile: comps/llms/text-generation/vllm/langchain/dependency/Dockerfile.intel_hpu
-    image: ${REGISTRY:-opea}/llm-vllm-hpu:${TAG:-latest}
   llm-vllm-ray:
     build:
       dockerfile: comps/llms/text-generation/vllm/ray/Dockerfile

comps/llms/text-generation/vllm/langchain/dependency/build_docker_vllm.sh

@@ -30,7 +30,11 @@ fi
 
 # Build the docker image for vLLM based on the hardware mode
 if [ "$hw_mode" = "hpu" ]; then
-    docker build -f Dockerfile.intel_hpu -t opea/vllm:hpu --shm-size=128g . --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy
+    git clone https://github.com/HabanaAI/vllm-fork.git
+    cd ./vllm-fork/
+    docker build -f Dockerfile.hpu -t opea/vllm:hpu --shm-size=128g . --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy
+    cd ..
+    rm -rf vllm-fork
 else
     git clone https://github.com/vllm-project/vllm.git
     cd ./vllm/

comps/llms/text-generation/vllm/langchain/dependency/launch_vllm_service.sh

@@ -38,7 +38,7 @@ volume=$PWD/data
 
 # Build the Docker run command based on hardware mode
 if [ "$hw_mode" = "hpu" ]; then
-    docker run -d --rm --runtime=habana --name="vllm-service" -p $port_number:80 -e HABANA_VISIBLE_DEVICES=all -e OMPI_MCA_btl_vader_single_copy_mechanism=none --cap-add=sys_nice --ipc=host -e HTTPS_PROXY=$https_proxy -e HTTP_PROXY=$https_proxy -e HF_TOKEN=${HF_TOKEN} opea/vllm:hpu /bin/bash -c "export VLLM_CPU_KVCACHE_SPACE=40 && python3 -m vllm.entrypoints.openai.api_server --enforce-eager --model $model_name  --tensor-parallel-size $parallel_number --host 0.0.0.0 --port 80 --block-size $block_size --max-num-seqs  $max_num_seqs --max-seq_len-to-capture $max_seq_len_to_capture "
+    docker run -d --rm --runtime=habana --name="vllm-service" -p $port_number:80 -e HABANA_VISIBLE_DEVICES=all -e OMPI_MCA_btl_vader_single_copy_mechanism=none --cap-add=sys_nice --ipc=host -e HTTPS_PROXY=$https_proxy -e HTTP_PROXY=$https_proxy -e HF_TOKEN=${HF_TOKEN} opea/vllm:hpu --enforce-eager --model $model_name  --tensor-parallel-size $parallel_number --host 0.0.0.0 --port 80 --block-size $block_size --max-num-seqs  $max_num_seqs --max-seq_len-to-capture $max_seq_len_to_capture
 else
     docker run -d --rm --name="vllm-service" -p $port_number:80 --network=host -v $volume:/data -e HTTPS_PROXY=$https_proxy -e HTTP_PROXY=$https_proxy -e HF_TOKEN=${HF_TOKEN} -e VLLM_CPU_KVCACHE_SPACE=40 opea/vllm:cpu --model $model_name --host 0.0.0.0 --port 80
 fi

comps/llms/text-generation/vllm/langchain/docker_compose_llm.yaml

@@ -23,7 +23,7 @@ services:
     cap_add:
       - SYS_NICE
     ipc: host
-    command: /bin/bash -c "export VLLM_CPU_KVCACHE_SPACE=40 && python3 -m vllm.entrypoints.openai.api_server --enforce-eager --model $LLM_MODEL --tensor-parallel-size 1 --host 0.0.0.0 --port 80"
+    command: --enforce-eager --model $LLM_MODEL --tensor-parallel-size 1 --host 0.0.0.0 --port 80
   llm:
     image: opea/llm-vllm:latest
     container_name: llm-vllm-gaudi-server

comps/llms/text-generation/vllm/llama_index/dependency/build_docker_vllm.sh

@@ -30,7 +30,11 @@ fi
 
 # Build the docker image for vLLM based on the hardware mode
 if [ "$hw_mode" = "hpu" ]; then
-    docker build -f docker/Dockerfile.intel_hpu -t opea/vllm:hpu --shm-size=128g . --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy
+    git clone https://github.com/HabanaAI/vllm-fork.git
+    cd ./vllm-fork/
+    docker build -f Dockerfile.hpu -t opea/vllm:hpu --shm-size=128g . --build-arg https_proxy=$https_proxy --build-arg http_proxy=$http_proxy
+    cd ..
+    rm -rf vllm-fork
 else
     git clone https://github.com/vllm-project/vllm.git
     cd ./vllm/

comps/llms/text-generation/vllm/llama_index/dependency/launch_vllm_service.sh

@@ -38,7 +38,7 @@ volume=$PWD/data
 
 # Build the Docker run command based on hardware mode
 if [ "$hw_mode" = "hpu" ]; then
-    docker run -d --rm --runtime=habana --name="vllm-service" -p $port_number:80 -e HABANA_VISIBLE_DEVICES=all -e OMPI_MCA_btl_vader_single_copy_mechanism=none --cap-add=sys_nice --ipc=host -e HTTPS_PROXY=$https_proxy -e HTTP_PROXY=$https_proxy -e HF_TOKEN=${HUGGINGFACEHUB_API_TOKEN} opea/vllm:hpu /bin/bash -c "export VLLM_CPU_KVCACHE_SPACE=40 && python3 -m vllm.entrypoints.openai.api_server --enforce-eager --model $model_name  --tensor-parallel-size $parallel_number --host 0.0.0.0 --port 80 --block-size $block_size --max-num-seqs  $max_num_seqs --max-seq_len-to-capture $max_seq_len_to_capture "
+    docker run -d --rm --runtime=habana --name="vllm-service" -p $port_number:80 -e HABANA_VISIBLE_DEVICES=all -e OMPI_MCA_btl_vader_single_copy_mechanism=none --cap-add=sys_nice --ipc=host -e HTTPS_PROXY=$https_proxy -e HTTP_PROXY=$https_proxy -e HF_TOKEN=${HUGGINGFACEHUB_API_TOKEN} opea/vllm:hpu --enforce-eager --model $model_name  --tensor-parallel-size $parallel_number --host 0.0.0.0 --port 80 --block-size $block_size --max-num-seqs  $max_num_seqs --max-seq_len-to-capture $max_seq_len_to_capture
 else
     docker run -d --rm --name="vllm-service" -p $port_number:80 --network=host -v $volume:/data -e HTTPS_PROXY=$https_proxy -e HTTP_PROXY=$https_proxy -e HF_TOKEN=${HUGGINGFACEHUB_API_TOKEN} -e VLLM_CPU_KVCACHE_SPACE=40 opea/vllm:cpu --model $model_name --host 0.0.0.0 --port 80
 fi

comps/llms/text-generation/vllm/llama_index/docker_compose_llm.yaml

@@ -23,7 +23,7 @@ services:
     cap_add:
       - SYS_NICE
     ipc: host
-    command: /bin/bash -c "export VLLM_CPU_KVCACHE_SPACE=40 && python3 -m vllm.entrypoints.openai.api_server --enforce-eager --model $LLM_MODEL --tensor-parallel-size 1 --host 0.0.0.0 --port 80"
+    command: --enforce-eager --model $LLM_MODEL --tensor-parallel-size 1 --host 0.0.0.0 --port 80
   llm:
     image: opea/llm-vllm-llamaindex:latest
     container_name: llm-vllm-gaudi-server

comps/llms/text-generation/vllm/llama_index/llm.py

@@ -66,7 +66,8 @@ async def stream_generator():
 
         return StreamingResponse(stream_generator(), media_type="text/event-stream")
     else:
-        response = await llm.acomplete(input.query).text
+        response = await llm.acomplete(input.query)
+        response = response.text
         if logflag:
             logger.info(response)
         return GeneratedDoc(text=response, prompt=input.query)

tests/llms/test_llms_text-generation_vllm_langchain_on_intel_hpu.sh

@@ -8,12 +8,11 @@ WORKPATH=$(dirname "$PWD")
 ip_address=$(hostname -I | awk '{print $1}')
 
 function build_docker_images() {
-    ## Build VLLM Ray docker
-    cd $WORKPATH/comps/llms/text-generation/vllm/langchain/dependency
-    docker build \
-        -f Dockerfile.intel_hpu \
-        --no-cache -t opea/vllm-hpu:comps \
-        --shm-size=128g .
+    ## Build VLLM docker
+    cd $WORKPATH
+    git clone https://github.com/HabanaAI/vllm-fork.git
+    cd vllm-fork/
+    docker build --no-cache -f Dockerfile.hpu -t opea/vllm-hpu:comps --shm-size=128g .
     if [ $? -ne 0 ]; then
         echo "opea/vllm-hpu built fail"
         exit 1
@@ -48,7 +47,7 @@ function start_service() {
         --ipc=host \
         -e HF_TOKEN=${HUGGINGFACEHUB_API_TOKEN} \
         opea/vllm-hpu:comps \
-        /bin/bash -c "export VLLM_CPU_KVCACHE_SPACE=40 && python3 -m vllm.entrypoints.openai.api_server --enforce-eager --model $LLM_MODEL  --tensor-parallel-size 1 --host 0.0.0.0 --port 80 --block-size 128 --max-num-seqs 256 --max-seq_len-to-capture 2048"
+        --enforce-eager --model $LLM_MODEL  --tensor-parallel-size 1 --host 0.0.0.0 --port 80 --block-size 128 --max-num-seqs 256 --max-seq_len-to-capture 2048
 
     export vLLM_ENDPOINT="http://${ip_address}:${port_number}"
     docker run -d --rm \
@@ -65,7 +64,7 @@ function start_service() {
     until [[ "$n" -ge 120 ]] || [[ $ready == true ]]; do
         docker logs test-comps-vllm-service > ${WORKPATH}/tests/test-comps-vllm-service.log
         n=$((n+1))
-        if grep -q Connected ${WORKPATH}/tests/test-comps-vllm-service.log; then
+        if grep -q throughput ${WORKPATH}/tests/test-comps-vllm-service.log; then
             break
         fi
         sleep 5s

tests/llms/test_llms_text-generation_vllm_llamaindex_on_intel_hpu.sh

@@ -8,12 +8,11 @@ WORKPATH=$(dirname "$PWD")
 ip_address=$(hostname -I | awk '{print $1}')
 
 function build_docker_images() {
-    ## Build VLLM Ray docker
-    cd $WORKPATH/comps/llms/text-generation/vllm/llama_index/dependency
-    docker build \
-        -f Dockerfile.intel_hpu \
-        --no-cache -t opea/vllm-hpu:comps \
-        --shm-size=128g .
+    ## Build VLLM docker
+    cd $WORKPATH
+    git clone https://github.com/HabanaAI/vllm-fork.git
+    cd vllm-fork/
+    docker build --no-cache -f Dockerfile.hpu -t opea/vllm-hpu:comps --shm-size=128g .
     if [ $? -ne 0 ]; then
         echo "opea/vllm-hpu built fail"
         exit 1
@@ -48,7 +47,7 @@ function start_service() {
         --ipc=host \
         -e HF_TOKEN=${HUGGINGFACEHUB_API_TOKEN} \
         opea/vllm-hpu:comps \
-        /bin/bash -c "export VLLM_CPU_KVCACHE_SPACE=40 && python3 -m vllm.entrypoints.openai.api_server --enforce-eager --model $LLM_MODEL  --tensor-parallel-size 1 --host 0.0.0.0 --port 80 --block-size 128 --max-num-seqs 256 --max-seq_len-to-capture 2048"
+        --enforce-eager --model $LLM_MODEL  --tensor-parallel-size 1 --host 0.0.0.0 --port 80 --block-size 128 --max-num-seqs 256 --max-seq_len-to-capture 2048
 
     export vLLM_ENDPOINT="http://${ip_address}:${port_number}"
     docker run -d --rm \
@@ -65,7 +64,7 @@ function start_service() {
     until [[ "$n" -ge 120 ]] || [[ $ready == true ]]; do
         docker logs test-comps-vllm-service > ${WORKPATH}/tests/test-comps-vllm-service.log
         n=$((n+1))
-        if grep -q Connected ${WORKPATH}/tests/test-comps-vllm-service.log; then
+        if grep -q throughput ${WORKPATH}/tests/test-comps-vllm-service.log; then
             break
         fi
         sleep 5s