Update Dockerfile

Dockerfile

@@ -1,42 +1,34 @@
-FROM python:3.14-slim
-
-# Set environment variables for caching to a writable directory
-# These will be used by huggingface libraries to store models and other assets.
-ENV HF_HOME=/data/huggingface
-ENV TORCH_HOME=/data/torch
-ENV TRANSFORMERS_CACHE=/data/huggingface/hub
-ENV HF_DATASETS_CACHE=/data/huggingface/datasets
-
-# Create the cache directories and make them writable by any user
-RUN mkdir -p $TRANSFORMERS_CACHE $HF_DATASETS_CACHE $TORCH_HOME && \
-    chmod -R 777 /data
-
-# System deps for soundfile, audio encoding, and uroman
+FROM python:3.11-slim
+
+# Set environment variables for writable caches on Hugging Face Spaces
+ENV DEBIAN_FRONTEND=noninteractive \
+    PIP_NO_CACHE_DIR=1 \
+    PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    PORT=7860 \
+    # Point all caches to the writable /data volume
+    HOME=/data \
+    XDG_CACHE_HOME=/data/.cache \
+    HF_HOME=/data/.cache/huggingface \
+    HUGGINGFACE_HUB_CACHE=/data/.cache/huggingface/hub \
+    TRANSFORMERS_CACHE=/data/transformers \
+    TORCH_HOME=/data/torch \
+    HF_HUB_DISABLE_TELEMETRY=1
+
+# System dependencies for audio processing
 RUN apt-get update && apt-get install -y --no-install-recommends \
-    libsndfile1 ffmpeg git perl && \
-    rm -rf /var/lib/apt/lists/*
+    libsndfile1 \
+  && rm -rf /var/lib/apt/lists/*
 
-# Clone and install uroman from source so the tokenizer can find it
-RUN git clone https://github.com/isi-nlp/uroman.git /opt/uroman && \
-    ln -s /opt/uroman/bin/uroman.pl /usr/local/bin/uroman
+# Create the cache directories and make them writable
+RUN mkdir -p /data/.cache/huggingface/hub /data/transformers /data/torch && chmod -R 777 /data
 
+# Set up the app
 WORKDIR /app
-
-# Install Python deps
 COPY backend/requirements.txt /app/requirements.txt
-RUN pip install --no-cache-dir --upgrade pip && \
-    pip install --no-cache-dir -r /app/requirements.txt
+RUN pip install -r /app/requirements.txt
 
-# Copy app code
 COPY backend/ /app/
 
-# Optional: pre-cache the MMS model at build time to speed up first request
-# This now uses the cache directory defined by the ENV variables.
-RUN python -c "from transformers import VitsModel, AutoTokenizer; model_id = 'facebook/mms-tts-amh'; VitsModel.from_pretrained(model_id); AutoTokenizer.from_pretrained(model_id); print('Pre-cached MMS model')"
-
-# Spaces expects the app to listen on $PORT (default 7860)
-ENV PORT=7860
 EXPOSE 7860
-
-# Start the Flask app via gunicorn
-CMD ["gunicorn", "app:app", "--bind", "0.0.0.0:7860", "--workers", "1", "--threads", "4", "--timeout", "300"]
+CMD ["gunicorn", "-w", "2", "-k", "gthread", "-b", "0.0.0.0:${PORT}", "app:app"]