Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,128 @@
+---
+language: en
+tags:
+- weather-forecasting
+- diffusion-models
+- rectified-flow
+- meteorology
+- pytorch
+- deep-learning
+license: mit
+datasets:
+- meteolibre
+---
+
+# MeteoLibre Rectified Flow Model
+
+This is a rectified flow diffusion model trained for meteorological data forecasting using the MeteoLibre dataset. The model uses a 3D U-Net architecture with FiLM conditioning for efficient weather pattern generation.
+
+## Model Description
+
+- **Model type**: Rectified Flow Diffusion Model
+- **Architecture**: 3D DC-AE U-Net with FiLM conditioning
+- **Input**: Meteorological data patches (12 channels, 3D spatio-temporal)
+- **Output**: Generated weather forecast data
+- **Training data**: MeteoLibre meteorological dataset
+- **Language(s)**: Python
+- **License**: MIT
+
+## Intended Use
+
+This model is designed for:
+- Weather pattern generation and forecasting
+- Meteorological data augmentation
+- Research in atmospheric science and weather prediction
+- Educational purposes in machine learning for climate modeling
+
+## Model Architecture
+
+The model consists of:
+- **UNet_DCAE_3D**: 3D convolutional U-Net with encoder-decoder architecture
+- **FiLM Conditioning**: Feature-wise linear modulation for temporal context
+- **Rectified Flow**: Efficient generative modeling approach
+- **Input channels**: 12 (meteorological variables)
+- **Output channels**: 12 (forecast variables)
+- **Features**: [64, 128, 256] channel progression
+- **Context frames**: 4 (temporal conditioning)
+
+## Training
+
+The model was trained using:
+- **Framework**: PyTorch with Hugging Face Accelerate
+- **Optimizer**: Adam (lr=5e-4)
+- **Batch size**: 64
+- **Epochs**: 200
+- **Precision**: Mixed precision (bf16)
+- **Distributed training**: Multi-GPU support
+
+## Usage
+
+### Loading the Model
+
+```python
+from safetensors.torch import load_file
+import torch
+from meteolibre_model.models.dc_3dunet_film import UNet_DCAE_3D
+
+# Load model weights
+state_dict = load_file("epoch_141_rectified_flow.safetensors")
+
+# Create model
+model = UNet_DCAE_3D(
+    in_channels=12,
+    out_channels=12,
+    features=[64, 128, 256],
+    context_dim=4,
+    context_frames=4,
+    num_additional_resnet_blocks=2
+)
+
+model.load_state_dict(state_dict)
+model.eval()
+```
+
+### Inference
+
+```python
+# Example inference code
+with torch.no_grad():
+    generated_data = model(input_batch)
+```
+
+## Performance
+
+The model checkpoints are saved at regular intervals:
+- epoch_1_rectified_flow.safetensors through epoch_141_rectified_flow.safetensors
+- Best performing checkpoints available for different training stages
+
+## Limitations
+
+- Model trained on specific meteorological dataset
+- May not generalize to all weather patterns or regions
+- Requires significant computational resources for inference
+- Temporal context limited to 4 frames
+
+## Ethical Considerations
+
+- Weather forecasting models should be used responsibly
+- Consider environmental impact of computational requirements
+- Validate predictions against ground truth data
+- Not intended for critical decision-making without human oversight
+
+## Citation
+
+If you use this model in your research, please cite:
+
+```bibtex
+@misc{meteolibre-rectified-flow,
+  title={MeteoLibre Rectified Flow Weather Forecasting Model},
+  author={MeteoLibre Development Team},
+  year={2025},
+  publisher={Hugging Face},
+  url={https://huggingface.co/meteolibre-dev/meteolibre-rectified-flow}
+}
+```
+
+## Contact
+
+For questions or issues, please open an issue on the [MeteoLibre GitHub repository](https://github.com/meteolibre-dev/meteolibre_model).