push

Dockerfile

@@ -1,39 +1,56 @@
-# syntax=docker/dockerfile:1
-
+# Use a lightweight Python base
 FROM python:3.10-slim
 
-ENV PYTHONDONTWRITEBYTECODE=1 \
+# Prevent interactive prompts & speed up Python
+ENV DEBIAN_FRONTEND=noninteractive \
     PYTHONUNBUFFERED=1 \
+    PYTHONDONTWRITEBYTECODE=1 \
     PIP_NO_CACHE_DIR=1 \
-    HF_HOME=/app/hf-home \
-    HUGGINGFACE_HUB_CACHE=/app/hf-home \
-    TRANSFORMERS_CACHE=/app/hf-home/transformers \
-    TOKENIZERS_PARALLELISM=false \
-    PORT=8000
+    TOKENIZERS_PARALLELISM=false
 
-WORKDIR /app
+# Set work directory
+WORKDIR /code
 
-# System deps (for scientific libs and uvicorn)
+# Install system dependencies
 RUN apt-get update && apt-get install -y --no-install-recommends \
     build-essential \
     git \
+    curl \
+    libopenblas-dev \
+    libomp-dev \
     && rm -rf /var/lib/apt/lists/*
 
-# Ensure Hugging Face cache dirs exist and are writable
-RUN mkdir -p /app/hf-home/transformers && chmod -R 777 /app/hf-home
+# Copy requirements first (for Docker caching)
+COPY requirements.txt .
+
+# Install Python dependencies
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Hugging Face tools
+RUN pip install --no-cache-dir huggingface-hub accelerate
+
+# Set Hugging Face cache inside container (persistent, not /tmp)
+ENV HF_HOME=/models/huggingface
+ENV TRANSFORMERS_CACHE=/models/huggingface
+ENV HUGGINGFACE_HUB_CACHE=/models/huggingface
+ENV HF_HUB_CACHE=/models/huggingface
+
+# Create cache dir
+RUN mkdir -p /models/huggingface
 
-# Copy only requirements first for better layer caching
-COPY requirements.txt ./
+# Pre-download model at build time (NLLB model)
+RUN python -c "from huggingface_hub import snapshot_download; snapshot_download(repo_id='drrobot9/nllb-ig-yo-ha-finetuned')"
 
-RUN pip install --upgrade pip && \
-    pip install -r requirements.txt
+# Preload tokenizer (avoid runtime delays)
+RUN python -c "from transformers import AutoTokenizer; AutoTokenizer.from_pretrained('drrobot9/nllb-ig-yo-ha-finetuned', use_fast=True)"
 
-# Copy the rest of the application
+# Copy project files
 COPY . .
 
-EXPOSE 8000
+# Expose FastAPI port
+EXPOSE 7860
 
-# Default command: serve FastAPI via uvicorn, binding to ${PORT}
-CMD ["sh", "-c", "uvicorn multilingual.multi:app --host 0.0.0.0 --port ${PORT:-8000} --lifespan off --timeout-keep-alive 75"]
+# Run FastAPI app with uvicorn (2 workers for better concurrency)
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860", "--workers", "2"]
 
 

main.py

@@ -1,10 +1,11 @@
 import os
 import uvicorn
+from multilingual.multi import app
 
 
 if __name__ == "__main__":
     host = os.getenv("HOST", "0.0.0.0")
-    port = int(os.getenv("PORT", "8000"))
+    port = int(os.getenv("PORT", "7860"))
     uvicorn.run("multilingual.multi:app", host=host, port=port, reload=False)