Refactor reranking (#1113)

Signed-off-by: WenjiaoYue <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>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: ZePan110 <ze.pan@intel.com>

comps/finetuning/README.md

@@ -244,7 +244,7 @@ curl http://${your_ip}:8015/v1/finetune/list_checkpoints -X POST -H "Content-Typ
 
 ### 3.4 Leverage fine-tuned model
 
-After fine-tuning job is done, fine-tuned model can be chosen from listed checkpoints, then the fine-tuned model can be used in other microservices. For example, fine-tuned reranking model can be used in [reranks](../reranks/src/README.md) microservice by assign its path to the environment variable `RERANK_MODEL_ID`, fine-tuned embedding model can be used in [embeddings](../embeddings/src/README.md) microservice by assign its path to the environment variable `model`, LLMs after instruction tuning can be used in [llms](../llms/src/text-generation/README.md) microservice by assign its path to the environment variable `your_hf_llm_model`.
+After fine-tuning job is done, fine-tuned model can be chosen from listed checkpoints, then the fine-tuned model can be used in other microservices. For example, fine-tuned reranking model can be used in [rerankings](../rerankings/src/README.md) microservice by assign its path to the environment variable `RERANK_MODEL_ID`, fine-tuned embedding model can be used in [embeddings](../embeddings/src/README.md) microservice by assign its path to the environment variable `model`, LLMs after instruction tuning can be used in [llms](../llms/src/text-generation/README.md) microservice by assign its path to the environment variable `your_hf_llm_model`.
 
 ## 🚀4. Descriptions for Finetuning parameters
 

comps/lvms/llama-vision/transformers_generation_utils.py

@@ -4238,7 +4238,7 @@ def _ranking_fast(
     alpha: float,
     beam_width: int,
 ) -> torch.FloatTensor:
-    """Reranks the top_k candidates based on a degeneration penalty (cosine similarity with previous tokens), as described
+    """Rerankings the top_k candidates based on a degeneration penalty (cosine similarity with previous tokens), as described
     in the paper "A Contrastive Framework for Neural Text Generation".
 
     Returns the index of the best candidate for each

comps/rerankings/deployment/docker_compose/rerank_tei.yaml

@@ -29,6 +29,7 @@ services:
       http_proxy: ${http_proxy}
       https_proxy: ${https_proxy}
       TEI_RERANKING_ENDPOINT: ${TEI_RERANKING_ENDPOINT}
+      RERANK_COMPONENT_NAME: "OPEA_TEI_RERANKING"
       HF_TOKEN: ${HF_TOKEN}
     depends_on:
       tei_reranking_service:

comps/rerankings/deployment/docker_compose/rerank_videoqna.yaml

@@ -0,0 +1,22 @@
+# Copyright (C) 2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+services:
+  reranking:
+    image: opea/reranking:latest
+    container_name: reranking-videoqna-server
+    ports:
+      - "8000:8000"
+    ipc: host
+    environment:
+      no_proxy: ${no_proxy}
+      http_proxy: ${http_proxy}
+      https_proxy: ${https_proxy}
+      CHUNK_DURATION: ${CHUNK_DURATION}
+      FILE_SERVER_ENDPOINT: ${FILE_SERVER_ENDPOINT}
+      RERANK_COMPONENT_NAME: "OPEA_VIDEO_RERANKING"
+    restart: unless-stopped
+
+networks:
+  default:
+    driver: bridge

comps/rerankings/src/Dockerfile

@@ -0,0 +1,46 @@
+# Copyright (C) 2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+FROM python:3.11-slim
+
+ENV LANG=C.UTF-8
+
+ARG ARCH="cpu"
+ARG SERVICE="all"
+
+RUN apt-get update -y && apt-get install -y --no-install-recommends --fix-missing \
+    git \
+    libgl1-mesa-glx \
+    libjemalloc-dev
+
+RUN useradd -m -s /bin/bash user && \
+    mkdir -p /home/user && \
+    chown -R user /home/user/
+
+USER user
+
+COPY comps /home/user/comps
+
+RUN if [ ${ARCH} = "cpu" ]; then \
+    pip install --no-cache-dir torch --index-url https://download.pytorch.org/whl/cpu; \
+fi && \
+if [ ${SERVICE} = "videoqna" ]; then \
+    pip install --no-cache-dir --upgrade pip setuptools && \
+    pip install --no-cache-dir -r /home/user/comps/rerankings/src/requirements_videoqna.txt; \
+elif [ ${SERVICE} = "all" ]; then \
+    git clone https://github.com/IntelLabs/fastRAG.git /home/user/fastRAG && \
+    cd /home/user/fastRAG && \
+    pip install --no-cache-dir --upgrade pip && \
+    pip install --no-cache-dir . && \
+    pip install --no-cache-dir .[intel] && \
+    pip install --no-cache-dir -r /home/user/comps/rerankings/src/requirements_videoqna.txt; \
+fi && \
+pip install --no-cache-dir --upgrade pip setuptools && \
+pip install --no-cache-dir -r /home/user/comps/rerankings/src/requirements.txt;
+
+
+ENV PYTHONPATH=$PYTHONPATH:/home/user
+
+WORKDIR /home/user/comps/rerankings/src
+
+ENTRYPOINT ["python", "opea_reranking_microservice.py"]

comps/rerankings/src/integrations/tei.py

@@ -18,7 +18,7 @@ from comps.cores.proto.api_protocol import (
     RerankingResponseData,
 )
 
-logger = CustomLogger("reranking_tei")
+logger = CustomLogger("tei_reranking")
 logflag = os.getenv("LOGFLAG", False)
 
 # Environment variables
@@ -27,8 +27,8 @@ CLIENTID = os.getenv("CLIENTID")
 CLIENT_SECRET = os.getenv("CLIENT_SECRET")
 
 
-@OpeaComponentRegistry.register("OPEA_RERANK_TEI")
-class OPEATEIReranking(OpeaComponent):
+@OpeaComponentRegistry.register("OPEA_TEI_RERANKING")
+class OpeaTEIReranking(OpeaComponent):
     """A specialized reranking component derived from OpeaComponent for TEI reranking services.
 
     Attributes:

comps/rerankings/src/integrations/videoqna.py

@@ -0,0 +1,121 @@
+# Copyright (C) 2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import logging
+import os
+import re
+
+from fastapi import HTTPException
+
+from comps import CustomLogger, LVMVideoDoc, OpeaComponentRegistry, SearchedMultimodalDoc, ServiceType
+from comps.cores.common.component import OpeaComponent
+
+logger = CustomLogger("video_reranking")
+logflag = os.getenv("LOGFLAG", False)
+
+chunk_duration = os.getenv("CHUNK_DURATION", "10") or "10"
+chunk_duration = float(chunk_duration) if chunk_duration.isdigit() else 10.0
+
+file_server_endpoint = os.getenv("FILE_SERVER_ENDPOINT") or "http://0.0.0.0:6005"
+
+logging.basicConfig(
+    level=logging.INFO, format="%(levelname)s:     [%(asctime)s] %(message)s", datefmt="%d/%m/%Y %I:%M:%S"
+)
+
+
+def get_top_doc(top_n, videos) -> list:
+    hit_score = {}
+    if videos is None:
+        return None
+    for video_name in videos:
+        try:
+            if video_name not in hit_score.keys():
+                hit_score[video_name] = 0
+            hit_score[video_name] += 1
+        except KeyError as r:
+            logging.info(f"no video name {r}")
+
+    x = dict(sorted(hit_score.items(), key=lambda item: -item[1]))  # sorted dict of video name and score
+    top_n_names = list(x.keys())[:top_n]
+    logging.info(f"top docs = {x}")
+    logging.info(f"top n docs names = {top_n_names}")
+
+    return top_n_names
+
+
+def find_timestamp_from_video(metadata_list, video):
+    return next(
+        (metadata["timestamp"] for metadata in metadata_list if metadata["video"] == video),
+        None,
+    )
+
+
+def format_video_name(video_name):
+    # Check for an existing file extension
+    match = re.search(r"\.(\w+)$", video_name)
+
+    if match:
+        extension = match.group(1)
+        # If the extension is not 'mp4', raise an error
+        if extension != "mp4":
+            raise ValueError(f"Invalid file extension: .{extension}. Only '.mp4' is allowed.")
+
+    # Use regex to remove any suffix after the base name (e.g., '_interval_0', etc.)
+    base_name = re.sub(r"(_interval_\d+)?(\.mp4)?$", "", video_name)
+
+    # Add the '.mp4' extension
+    formatted_name = f"{base_name}.mp4"
+
+    return formatted_name
+
+
+@OpeaComponentRegistry.register("OPEA_VIDEO_RERANKING")
+class OpeaVideoReranking(OpeaComponent):
+    """A specialized reranking component derived from OpeaComponent for OPEA Video native reranking services."""
+
+    def __init__(self, name: str, description: str, config: dict = None):
+        super().__init__(name, ServiceType.RERANK.name.lower(), description, config)
+
+    async def invoke(self, input: SearchedMultimodalDoc) -> LVMVideoDoc:
+        """Invokes the reranking service to generate reranking for the provided input.
+
+        Args:
+            input (SearchedMultimodalDoc): The input in OpenAI reranking format.
+
+        Returns:
+            LVMVideoDoc: The response in OpenAI reranking format.
+        """
+        try:
+            # get top video name from metadata
+            video_names = [meta["video"] for meta in input.metadata]
+            top_video_names = get_top_doc(input.top_n, video_names)
+
+            # only use the first top video
+            timestamp = find_timestamp_from_video(input.metadata, top_video_names[0])
+            formatted_video_name = format_video_name(top_video_names[0])
+            video_url = f"{file_server_endpoint.rstrip('/')}/{formatted_video_name}"
+
+            result = LVMVideoDoc(
+                video_url=video_url,
+                prompt=input.initial_query,
+                chunk_start=timestamp,
+                chunk_duration=float(chunk_duration),
+                max_new_tokens=512,
+            )
+        except ValueError as e:
+            raise HTTPException(status_code=400, detail=str(e))
+        except Exception as e:
+            logging.error(f"Unexpected error in reranking: {str(e)}")
+            # Handle any other exceptions with a generic server error response
+            raise HTTPException(status_code=500, detail="An unexpected error occurred.")
+
+        return result
+
+    def check_health(self) -> bool:
+        """Checks the health of the reranking service.
+
+        Returns:
+            bool: True if the service is reachable and healthy, False otherwise.
+        """
+
+        return True

comps/rerankings/src/opea_reranking_microservice.py

@@ -5,7 +5,8 @@ import os
 import time
 from typing import Union
 
-from integrations.opea_tei import OPEATEIReranking
+from integrations.tei import OpeaTEIReranking
+from integrations.videoqna import OpeaVideoReranking
 
 from comps import (
     CustomLogger,
@@ -22,7 +23,7 @@ from comps.cores.proto.docarray import LLMParamsDoc, LVMVideoDoc, RerankedDoc, S
 logger = CustomLogger("opea_reranking_microservice")
 logflag = os.getenv("LOGFLAG", False)
 
-rerank_component_name = os.getenv("RERANK_COMPONENT_NAME", "OPEA_RERANK_TEI")
+rerank_component_name = os.getenv("RERANK_COMPONENT_NAME", "OPEA_TEI_RERANKING")
 # Initialize OpeaComponentLoader
 loader = OpeaComponentLoader(rerank_component_name, description=f"OPEA RERANK Component: {rerank_component_name}")
 

comps/rerankings/src/requirements_videoqna.txt

@@ -0,0 +1,7 @@
+datasets
+haystack-ai
+langchain  --extra-index-url https://download.pytorch.org/whl/cpu
+langchain_community  --extra-index-url https://download.pytorch.org/whl/cpu
+openai
+Pillow
+pydub

comps/reranks/src/Dockerfile

@@ -1,30 +0,0 @@
-# Copyright (C) 2024 Intel Corporation
-# SPDX-License-Identifier: Apache-2.0
-
-FROM python:3.11-slim
-
-ENV LANG=C.UTF-8
-
-ARG ARCH="cpu"
-
-RUN apt-get update -y && apt-get install -y --no-install-recommends --fix-missing \
-    libgl1-mesa-glx \
-    libjemalloc-dev
-
-RUN useradd -m -s /bin/bash user && \
-    mkdir -p /home/user && \
-    chown -R user /home/user/
-
-USER user
-
-COPY comps /home/user/comps
-
-RUN pip install --no-cache-dir --upgrade pip setuptools && \
-    if [ ${ARCH} = "cpu" ]; then pip install --no-cache-dir torch --index-url https://download.pytorch.org/whl/cpu; fi && \
-    pip install --no-cache-dir -r /home/user/comps/reranks/src/requirements.txt
-
-ENV PYTHONPATH=$PYTHONPATH:/home/user
-
-WORKDIR /home/user/comps/reranks/src
-
-ENTRYPOINT ["python", "opea_reranking_microservice.py"]

comps/text2sql/src/Dockerfile

@@ -30,4 +30,4 @@ ENV PYTHONPATH=$PYTHONPATH:/home/user
 
 WORKDIR /home/user/comps/text2sql/src/
 
-ENTRYPOINT ["python", "opea_text2sql_microservice.py"]
+ENTRYPOINT ["python", "opea_text2sql_microservice.py"]