Akshayuc16/QuPep-GenAI

IDR Energy-aware VAE (Sampler)

This repository is intended to host a trained checkpoint (model.pt) for the energy-aware VAE used to sample IDR conformational ensembles.

Files

• model.pt: PyTorch checkpoint produced by the training step in idr_quantum_genai_pipeline.py.
• inference.py: Inference-only loader + sampler utilities (no Qiskit dependencies).

How to sample locally

python -c "import torch; from inference import load_model_from_checkpoint, sample_ensemble_arrays; m=load_model_from_checkpoint('model.pt','cpu'); out=sample_ensemble_arrays(m,'DSKRETYS',n_samples=100); print(out['energy'][:5], out['angles'].shape)"

Notes

• Energies are model-predicted energies (learned surrogate), not an ab initio physics calculation.
• The public Space should cap n_samples to avoid timeouts.