e5-large-v2

Runtime error

App Files Files Community

limcheekin commited on Aug 19, 2023

Commit

b92d070

1 Parent(s): 990a127

feat: first import

Browse files

Files changed (9) hide show

Dockerfile +44 -0
README.md +11 -2
download.sh +9 -0
index.html +39 -0
open/__init__.py +0 -0
open/text/embeddings/server/__main__.py +37 -0
open/text/embeddings/server/app.py +114 -0
server-requirements.txt +5 -0
start_server.sh +3 -0

Dockerfile ADDED Viewed

	@@ -0,0 +1,44 @@

+FROM debian:bullseye-slim AS build-image
+ARG MODEL="BAAI/bge-large-en"
+ENV MODEL=${MODEL}
+COPY ./download.sh ./
+# Install build dependencies
+RUN apt-get update && \
+    apt-get install -y git-lfs
+RUN chmod +x *.sh && \
+    ./download.sh && \
+    rm *.sh
+# Stage 3 - final runtime image
+# Grab a fresh copy of the Python image
+FROM python:3.10-slim
+ARG MODEL="BAAI/bge-large-en"
+ENV MODEL=${MODEL}
+ENV NORMALIZE_EMBEDDINGS=1
+RUN mkdir -p ${MODEL} && mkdir -p open/text/embeddings
+COPY --from=build-image ${MODEL} ${MODEL}
+COPY open/text/embeddings ./open/text/embeddings
+COPY server-requirements.txt ./
+RUN pip install --no-cache-dir -r server-requirements.txt
+COPY ./start_server.sh ./
+COPY ./index.html ./
+# Make the server start script executable
+RUN chmod +x ./start_server.sh
+# Set environment variable for the host
+ENV HOST=0.0.0.0
+ENV PORT=7860
+# Expose a port for the server
+EXPOSE ${PORT}
+# Run the server start script
+CMD ["/bin/sh", "./start_server.sh"]

README.md CHANGED Viewed

@@ -1,10 +1,19 @@
 ---
-title: Bge Large En
 emoji: 👁
 colorFrom: gray
 colorTo: pink
 sdk: docker
 pinned: false
 ---
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

 ---
+title: BAAI/bge-large-en OpenAI API-Compatible Endpoint
 emoji: 👁
 colorFrom: gray
 colorTo: pink
 sdk: docker
+models:
+  - BAAI/bge-large-en
+tags:
+  - inference api
+  - openai-api compatible
+  - open-text-embeddings
+  - bge-large-en
 pinned: false
 ---
+# BAAI/bge-large-en OpenAI API-Compatible Endpoint
+Please refer to the [main screen](https://huggingface.co/spaces/limcheekin/bge-large-en) for more information.

download.sh ADDED Viewed

	@@ -0,0 +1,9 @@

+mkdir -p $MODEL
+git lfs install --skip-smudge
+git clone https://huggingface.co/$MODEL $MODEL
+cd $MODEL
+git lfs pull
+git lfs install --force
+rm -rf .git
+pwd
+ls -l

index.html ADDED Viewed

	@@ -0,0 +1,39 @@

+<!DOCTYPE html>
+<html>
+  <head>
+    <title>BAAI/bge-large-en OpenAI API-Compatible Endpoint</title>
+  </head>
+  <body>
+    <h1>BAAI/bge-large-en OpenAI API-Compatible Endpoint</h1>
+    <p>
+      With the utilization of the
+      <a href="https://github.com/limcheekin/open-text-embeddings"
+        >open-text-embeddings</a
+      >
+      package, we are excited to introduce the text embeddings model hosted in
+      the Hugging Face Docker Spaces, made accessible through an
+      OpenAI-compatible API. This space includes comprehensive API documentation
+      to facilitate seamless integration.
+    </p>
+    <ul>
+      <li>
+        The API endpoint:
+        <a href="https://limcheekin-bge-large-en.hf.space/v1"
+          >https://limcheekin-bge-large-en.hf.space/v1</a
+        >
+      </li>
+      <li>
+        The API doc:
+        <a href="https://limcheekin-bge-large-en.hf.space/docs"
+          >https://limcheekin-bge-large-en.hf.space/docs</a
+        >
+      </li>
+    </ul>
+    <p>
+      If you find this resource valuable, your support in the form of starring
+      the space would be greatly appreciated. Your engagement plays a vital role
+      in furthering the application for a community GPU grant, ultimately
+      enhancing the capabilities and accessibility of this space.
+    </p>
+  </body>
+</html>

open/__init__.py ADDED Viewed

File without changes

open/text/embeddings/server/__main__.py ADDED Viewed

	@@ -0,0 +1,37 @@

+"""FastAPI server for open-text-embeddings.
+To run this example:
+```bash
+pip install -r --no-cache-dir server-requirements.txt
+```
+Then run:
+```
+MODEL=intfloat/e5-large-v2 python -m open.text.embeddings.server
+```
+Then visit http://localhost:8000/docs to see the interactive API docs.
+"""
+import uvicorn
+from fastapi.responses import HTMLResponse
+from open.text.embeddings.server.app import create_app
+import os
+app = create_app()
+# Read the content of index.html once and store it in memory
+with open("index.html", "r") as f:
+    content = f.read()
+@app.get("/", response_class=HTMLResponse)
+async def read_items():
+    return content
+if __name__ == "__main__":
+    uvicorn.run(app,
+                host=os.environ["HOST"],
+                port=int(os.environ["PORT"])
+                )

open/text/embeddings/server/app.py ADDED Viewed

	@@ -0,0 +1,114 @@

+from typing import List, Optional, Union
+from starlette.concurrency import run_in_threadpool
+from fastapi import FastAPI, APIRouter
+from fastapi.middleware.cors import CORSMiddleware
+from pydantic import BaseModel, Field
+from langchain.embeddings import HuggingFaceEmbeddings
+from langchain.embeddings import HuggingFaceInstructEmbeddings
+from langchain.embeddings import HuggingFaceBgeEmbeddings
+import os
+router = APIRouter()
+DEFAULT_MODEL_NAME = "intfloat/e5-large-v2"
+E5_EMBED_INSTRUCTION = "passage: "
+E5_QUERY_INSTRUCTION = "query: "
+BGE_EN_QUERY_INSTRUCTION = "Represent this sentence for searching relevant passages: "
+BGE_ZH_QUERY_INSTRUCTION = "为这个句子生成表示以用于检索相关文章："
+def create_app():
+    app = FastAPI(
+        title="Open Text Embeddings API",
+        version="0.0.2",
+    )
+    app.add_middleware(
+        CORSMiddleware,
+        allow_origins=["*"],
+        allow_credentials=True,
+        allow_methods=["*"],
+        allow_headers=["*"],
+    )
+    app.include_router(router)
+    return app
+class CreateEmbeddingRequest(BaseModel):
+    model: Optional[str] = Field(
+        description="The model to use for generating embeddings.")
+    input: Union[str, List[str]] = Field(description="The input to embed.")
+    user: Optional[str]
+    class Config:
+        schema_extra = {
+            "example": {
+                "input": "The food was delicious and the waiter...",
+            }
+        }
+class Embedding(BaseModel):
+    embedding: List[float]
+class CreateEmbeddingResponse(BaseModel):
+    data: List[Embedding]
+embeddings = None
+def _create_embedding(
+    request: CreateEmbeddingRequest
+):
+    global embeddings
+    if embeddings is None:
+        if request.model and request.model != "text-embedding-ada-002":
+            model_name = request.model
+        else:
+            model_name = os.environ["MODEL"]
+        print("Loading model:", model_name)
+        encode_kwargs = {
+            "normalize_embeddings": bool(os.environ.get("NORMALIZE_EMBEDDINGS", ""))
+        }
+        print("encode_kwargs", encode_kwargs)
+        if "e5" in model_name:
+            embeddings = HuggingFaceInstructEmbeddings(model_name=model_name,
+                                                       embed_instruction=E5_EMBED_INSTRUCTION,
+                                                       query_instruction=E5_QUERY_INSTRUCTION,
+                                                       encode_kwargs=encode_kwargs)
+        elif model_name.startswith("BAAI/bge-") and model_name.endswith("-en"):
+            embeddings = HuggingFaceBgeEmbeddings(model_name=model_name,
+                                                    query_instruction=BGE_EN_QUERY_INSTRUCTION,
+                                                    encode_kwargs=encode_kwargs)
+        elif model_name.startswith("BAAI/bge-") and model_name.endswith("-zh"):
+            embeddings = HuggingFaceBgeEmbeddings(model_name=model_name,
+                                                    query_instruction=BGE_ZH_QUERY_INSTRUCTION,
+                                                    encode_kwargs=encode_kwargs)
+        else:
+            embeddings = HuggingFaceEmbeddings(
+                model_name=model_name, encode_kwargs=encode_kwargs)
+    if isinstance(request.input, str):
+        return CreateEmbeddingResponse(data=[Embedding(embedding=embeddings.embed_query(request.input))])
+    else:
+        data = [Embedding(embedding=embedding)
+                for embedding in embeddings.embed_documents(request.input)]
+        return CreateEmbeddingResponse(data=data)
+@router.post(
+    "/v1/embeddings",
+    response_model=CreateEmbeddingResponse,
+)
+async def create_embedding(
+    request: CreateEmbeddingRequest
+):
+    return _create_embedding(request)
+#    throw TypeError: 'CreateEmbeddingResponse' object is not callable?
+#    return await run_in_threadpool(
+#        _create_embedding(request)
+#    )

server-requirements.txt ADDED Viewed

	@@ -0,0 +1,5 @@

+fastapi
+sse-starlette
+sentence_transformers
+langchain
+uvicorn

start_server.sh ADDED Viewed

	@@ -0,0 +1,3 @@


1	+ #!/bin/sh
2	+
3	+ python -B -m open.text.embeddings.server