Upload folder using huggingface_hub

Dockerfile

@@ -1,27 +1,35 @@
 FROM python:3.10-slim
 
-WORKDIR /app
+# Set working directory
+WORKDIR /home/user/app
 
 # Install system dependencies
 RUN apt-get update && apt-get install -y \
     git \
+    curl \
     && rm -rf /var/lib/apt/lists/*
 
+# Create non-root user for HF Spaces
+RUN useradd -m -u 1000 user
+USER user
+ENV HOME=/home/user \
+    PATH=/home/user/.local/bin:$PATH
+
 # Copy requirements first for caching
-COPY requirements.txt .
+COPY --chown=user requirements.txt .
 
 # Install Python dependencies
-RUN pip install --no-cache-dir -r requirements.txt
+RUN pip install --no-cache-dir --upgrade pip && \
+    pip install --no-cache-dir -r requirements.txt
 
 # Copy application code
-COPY app.py .
+COPY --chown=user app.py .
 
 # Expose Streamlit port
 EXPOSE 7860
 
-# Set environment variables
-ENV STREAMLIT_SERVER_PORT=7860
-ENV STREAMLIT_SERVER_ADDRESS=0.0.0.0
+# Health check
+HEALTHCHECK CMD curl --fail http://localhost:7860/_stcore/health || exit 1
 
 # Run Streamlit
-CMD ["streamlit", "run", "app.py", "--server.port=7860", "--server.address=0.0.0.0"]
+ENTRYPOINT ["streamlit", "run", "app.py", "--server.port=7860", "--server.address=0.0.0.0", "--server.enableCORS=false", "--server.enableXsrfProtection=false"]

README.md

@@ -1,59 +1,33 @@
 ---
 title: WeatherFlow Training
-emoji: 🌤️
+emoji: 🌀
 colorFrom: blue
-colorTo: green
-sdk: gradio
-sdk_version: 4.44.0
-app_file: app.py
+colorTo: purple
+sdk: docker
 pinned: false
 license: mit
 ---
 
-# WeatherFlow Flow Matching Model Training
+# WeatherFlow Training
 
-This Space provides GPU-accelerated training for flow matching models used in weather prediction.
+Train flow matching models to predict wind fields from hurricane brightness temperature images.
 
 ## Features
 
-- **Real PyTorch Training**: Train actual flow matching models with GPU acceleration
-- **Physics-Informed Constraints**: Optional divergence and energy conservation losses
-- **Attention Mechanisms**: Multi-head attention for capturing spatial dependencies
-- **Model Upload**: Automatically uploads trained models to Hugging Face Hub
-- **Progress Tracking**: Real-time training progress and metrics
+- **Train Models**: Upload hurricane image pairs from Google Drive and train flow matching models
+- **Run Inference**: Use trained models to predict wind fields from new hurricane images
+- **Save to Hub**: Automatically save trained models to Hugging Face Hub
 
-## Usage
+## How to Use
 
-1. Configure your experiment parameters
-2. Enter your Hugging Face token (with write access)
-3. Click "Start Training"
-4. Wait for training to complete
-5. Find your trained model in your Hugging Face Hub
+### Training
+1. Prepare hurricane image pairs where left half = brightness temperature, right half = wind field
+2. Upload images to a Google Drive folder and make it publicly accessible
+3. Paste the Google Drive folder URL
+4. Configure model parameters and click "Start Training"
+5. The trained model will be saved to your Hugging Face Hub
 
-## API Usage
-
-You can also call this Space programmatically:
-
-```python
-from gradio_client import Client
-
-client = Client("your-username/weatherflow-training")
-result = client.predict(
-    experiment_name="My Experiment",
-    variables="z, t",
-    pressure_levels="500",
-    grid_lat=16,
-    grid_lon=32,
-    hidden_dim=96,
-    num_layers=3,
-    use_attention=True,
-    physics_informed=True,
-    epochs=50,
-    batch_size=8,
-    learning_rate=0.0005,
-    train_samples=64,
-    val_samples=16,
-    hf_token="your-token"
-)
-print(result)
-```
+### Inference
+1. Enter the HF Hub repository ID of your trained model
+2. Upload a hurricane brightness temperature image
+3. Click "Run Inference" to generate the predicted wind field