| # Model Files Information |
|
|
| ## Available Files |
| - normalization_stats.pt: Preprocessing statistics for ESM-2 embeddings |
| |
| ## Missing Files (Too Large for Hugging Face) |
| The following model files are too large for Hugging Face upload (>100MB limit): |
| |
| ### Large Model Files (Not Included) |
| - flowamp_demo_checkpoint.pth (~1.5GB): Complete model checkpoint |
| - compressor_demo.pth (~315MB): Compressor weights |
| - decompressor_demo.pth (~158MB): Decompressor weights |
| - flow_model_demo.pth (~54MB): Flow model weights |
| - apex/trained_models/* (~1GB total): Pre-trained Apex models |
|
|
| ### How to Get Model Files |
| 1. **Train Your Own**: Use the provided training scripts to train the model |
| 2. **Contact Author**: Request model files directly from the author |
| 3. **Alternative Storage**: Model files may be available on other platforms |
|
|
| ### Training Instructions |
| 1. Run the training scripts to generate your own model checkpoints |
| 2. Use amp_flow_training_single_gpu_full_data.py for single GPU training |
| 3. Use amp_flow_training_multi_gpu.py for multi-GPU training |
| 4. Models will be saved automatically during training |
|
|
| ### Quick Start |
| 1. Install dependencies: pip install -r requirements.txt |
| 2. Run usage_example.py to verify installation |
| 3. Train the model using provided scripts |
| 4. Use generate_amps.py for AMP generation |
|
|
