Update

Dockerfile

@@ -1,6 +1,7 @@
 FROM python:3.10
 ENV PIP_ROOT_USER_ACTION=ignore \
   TZ=Asia/Shanghai
+RUN ln -snf /usr/share/zoneinfo/$TZ /etc/localtime && echo $TZ > /etc/timezone
 
 WORKDIR /app
 COPY . .
@@ -20,8 +21,8 @@ RUN pip install --no-cache-dir --upgrade pip && \
 # CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "7860"]
 # CMD ["TRANSFORMERS_CACHE=./", "infinity_emb", "--model-name-or-path", "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2", "--port", "7860"]
 
-# x ["sh", "-c", "'FOO=BAR python app.py'"]
 # CMD ["python", "m3_server.py"]
+# x ["sh", "-c", "'FOO=BAR python m3_server.py'"]
+# CMD ["sh", "-c", "HF_HOME=/tmp/cache", "python", "m3_server.py"]
 
-# CMD ["sh", "start-m3-server.sh"]  # OK
-CMD ["sh", "-c", "HF_HOME=/tmp/cache", "python", "m3_server.py"]
+CMD ["sh", "start-m3-server.sh"]  # OK

m3_server.py

@@ -14,10 +14,12 @@ from FlagEmbedding import BGEM3FlagModel
 from pydantic import BaseModel
 from starlette.status import HTTP_504_GATEWAY_TIMEOUT
 
+_ = """
 Path("/tmp/cache").mkdir(exist_ok=True)
 os.environ["HF_HOME"] = "/tmp/cache"
 os.environ["TRANSFORMERS_CACHE"] = "/tmp/cache"
 # does not quite work
+# """
 
 batch_size = 2  # gpu batch_size in order of your available vram
 max_request = 10  # max request for future improvements on api calls / gpu batches (for now is pretty basic)
@@ -242,5 +244,5 @@ async def rerank(request: RerankRequest):
 
 if __name__ == "__main__":
     import uvicorn
-
+    print("started")
     uvicorn.run(app, host="0.0.0.0", port=port)