Merge pull request #88 from weaviate/introduce-multistage-dockerfile

Introduce multi stage Dockerfile

Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.11-slim
+FROM python:3.11-slim AS base_image
 
 WORKDIR /app
 
@@ -8,13 +8,23 @@ RUN pip install --upgrade pip setuptools
 COPY requirements.txt .
 RUN pip3 install -r requirements.txt
 
+FROM base_image AS download_model
+
+WORKDIR /app
+
 ARG TARGETARCH
 ARG MODEL_NAME
 ARG ONNX_RUNTIME
 ENV ONNX_CPU=${TARGETARCH}
+RUN mkdir nltk_data
 COPY download.py .
 RUN ./download.py
 
+FROM base_image AS t2v_transformers
+
+WORKDIR /app
+COPY --from=download_model /app/models /app/models
+COPY --from=download_model /app/nltk_data /app/nltk_data
 COPY . .
 
 ENTRYPOINT ["/bin/sh", "-c"]

download.py

@@ -16,6 +16,7 @@
 
 
 model_dir = './models/model'
+nltk_dir = './nltk_data'
 model_name = os.getenv('MODEL_NAME', None)
 force_automodel = os.getenv('FORCE_AUTOMODEL', False)
 if not model_name:
@@ -108,7 +109,7 @@ def download_model(model_name: str, model_dir: str):
         model.save_pretrained(model_dir)
         tokenizer.save_pretrained(model_dir)
 
-    nltk.download('punkt', download_dir='./nltk_data')
+    nltk.download('punkt', download_dir=nltk_dir)
 
 if onnx_runtime == "true":
     download_onnx_model(model_name, model_dir)