| from __future__ import annotations |
|
|
| from pathlib import Path |
|
|
| import torch |
|
|
| from . import __version__ |
| from .models import SSLModel |
|
|
|
|
| def save_checkpoint(path, model, method, config, epoch, metrics): |
| payload = { |
| "version": __version__, "method": method, "epoch": epoch, |
| "config": config, "metrics": metrics, "encoder": model.encoder.state_dict(), |
| "projector": model.projector.state_dict() if model.projector else None, |
| "rotation_head": model.rotation_head.state_dict() if model.rotation_head else None, |
| "decoder": model.decoder.state_dict() if model.decoder else None, |
| } |
| Path(path).parent.mkdir(parents=True, exist_ok=True) |
| torch.save(payload, path) |
|
|
|
|
| def load_checkpoint(path, device="cpu"): |
| payload = torch.load(path, map_location=device, weights_only=True) |
| model = SSLModel(payload["method"], pretrained=False) |
| model.encoder.load_state_dict(payload["encoder"]) |
| for name in ("projector", "rotation_head", "decoder"): |
| module, state = getattr(model, name), payload.get(name) |
| if module is not None and state is not None: |
| module.load_state_dict(state) |
| return model.to(device), payload |
|
|
|
|
