fix vllm output in chatqna (#1038)

Signed-off-by: Xinyao Wang <xinyao.wang@intel.com>

Author: XinyaoWa

ChatQnA/chatqna.py

@@ -166,7 +166,10 @@ def align_generator(self, gen, **kwargs):
         try:
             # sometimes yield empty chunk, do a fallback here
             json_data = json.loads(json_str)
-            if json_data["choices"][0]["finish_reason"] != "eos_token":
+            if (
+                json_data["choices"][0]["finish_reason"] != "eos_token"
+                and "content" in json_data["choices"][0]["delta"]
+            ):
                 yield f"data: {repr(json_data['choices'][0]['delta']['content'].encode('utf-8'))}\n\n"
         except Exception as e:
             yield f"data: {repr(json_str.encode('utf-8'))}\n\n"

ChatQnA/docker_image_build/build.yaml

@@ -83,12 +83,6 @@ services:
       dockerfile: comps/llms/text-generation/vllm/langchain/dependency/Dockerfile.intel_hpu
     extends: chatqna
     image: ${REGISTRY:-opea}/llm-vllm-hpu:${TAG:-latest}
-  llm-vllm-ray:
-    build:
-      context: GenAIComps
-      dockerfile: comps/llms/text-generation/vllm/ray/Dockerfile
-    extends: chatqna
-    image: ${REGISTRY:-opea}/llm-vllm-ray:${TAG:-latest}
   llm-vllm-ray-hpu:
     build:
       context: GenAIComps

docker_images_list.md

@@ -80,7 +80,6 @@ Take ChatQnA for example. ChatQnA is a chatbot application service based on the
 | [opea/llm-vllm-hpu](https://hub.docker.com/r/opea/llm-vllm-hpu)                                                     | [Link](https://github.com/opea-project/GenAIComps/blob/main/comps/llms/text-generation/vllm/langchain/dependency/Dockerfile.intel_hpu)   | The docker image exposed the OPEA LLM microservice upon vLLM docker image for use by GenAI apps on the Gaudi                                                                                                               |
 | [opea/llm-vllm-llamaindex](https://hub.docker.com/r/opea/llm-vllm-llamaindex)                                       | [Link](https://github.com/opea-project/GenAIComps/blob/main/comps/llms/text-generation/vllm/llama_index/Dockerfile)                      | This docker image exposes OPEA LLM microservices to the llamaindex framework's vLLM Docker image for use by GenAI applications                                                                                             |
 | [opea/llm-vllm-llamaindex-hpu](https://hub.docker.com/r/opea/llm-vllm-llamaindex-hpu)                               | [Link](https://github.com/opea-project/GenAIComps/blob/main/comps/llms/text-generation/vllm/llama_index/dependency/Dockerfile.intel_hpu) | This docker image exposes OPEA LLM microservices to the llamaindex framework's vLLM Docker image for use by GenAI applications on the gaudi                                                                                |
-| [opea/llm-vllm-ray](https://hub.docker.com/r/opea/llm-vllm-ray)                                                     | [Link](https://github.com/opea-project/GenAIComps/blob/main/comps/llms/text-generation/vllm/ray/Dockerfile)                              | The docker image exposes the OPEA LLM microservices Ray-based upon the vLLM Docker image for GenAI application use                                                                                                         |
 | [opea/llm-vllm-ray-hpu](https://hub.docker.com/r/opea/llm-vllm-ray-hpu)                                             | [Link](https://github.com/opea-project/GenAIComps/blob/main/comps/llms/text-generation/vllm/ray/dependency/Dockerfile)                   | The docker image exposes Ray-based OPEA LLM microservices upon the vLLM Docker image for use by GenAI applications on the Gaudi                                                                                            |
 | [opea/llava-hpu](https://hub.docker.com/r/opea/llava-hpu)                                                           | [Link](https://github.com/opea-project/GenAIComps/blob/main/comps/lvms/llava/dependency/Dockerfile.intel_hpu)                            | The docker image exposed the OPEA microservice running LLaVA as a large visual model (LVM) service for GenAI application use on the Gaudi                                                                                  |
 | [opea/lvm-tgi](https://hub.docker.com/r/opea/lvm-tgi)                                                               | [Link](https://github.com/opea-project/GenAIComps/blob/main/comps/lvms/tgi-llava/Dockerfile)                                             | This docker image is designed to build a large visual model (LVM) microservice using the HuggingFace Text Generation Inference(TGI) framework. The microservice accepts document input and generates a answer to question. |