Upload 2 files

Dockerfile

@@ -0,0 +1,24 @@
+FROM python:3.10-slim
+WORKDIR /app
+
+# The user will only upload app.zip and this Dockerfile
+COPY app.zip .
+
+# Install unzip, extract the application, and clean up the zip file
+# (also install build-essential for tokenizer dependencies)
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    unzip \
+    build-essential \
+    && unzip -o app.zip && rm app.zip \
+    && rm -rf /var/lib/apt/lists/*
+
+# Install dependencies from the extracted requirements.txt
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Pre-download the tokenizer during the Docker build (setting cache dir to /app/cache to ensure it persists across users)
+ENV HF_HOME=/app/cache
+RUN python -c "from transformers import AutoTokenizer; AutoTokenizer.from_pretrained('unsloth/Llama-3.2-3B-Instruct')"
+
+EXPOSE 8000
+# We use --preload so the master process downloads the tokenizer ONCE before forking the workers
+CMD ["gunicorn", "--preload", "--timeout", "120", "-w", "4", "-b", "0.0.0.0:8000", "app:app"]

app.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d754d77be3321297044f72c3b741d57033e9690e28bce9b7c1f32dccc75cfea3
+size 1110651