meybe add training again (read best od fine) and train again

``` # after config device checkpoint_path = "checkpoints/best_model.pt"

ckpt = torch.load(checkpoint_path, map_location=device, weights_only=False)

model = GuppyLM(mc).to(device) if "model_state_dict" in ckpt: model.load_state_dict(ckpt["model_state_dict"]) else: model.load_state_dict(ckpt)

start_step = ckpt.get("step", 0) print(f"Encore {start_step}") ```