| import os | |
| import torch | |
| def save_checkpoint(model, model_ema, optimizer, scheduler, epoch, work_dir=None): | |
| save_dir = os.path.join(work_dir, "checkpoint") | |
| save_states = { | |
| "epoch": epoch, | |
| "state_dict": model.state_dict(), | |
| "optimizer": optimizer.state_dict(), | |
| "scheduler": scheduler.state_dict(), | |
| } | |
| if model_ema != None: | |
| save_states.update({"state_dict_ema": model_ema.module.state_dict()}) | |
| if not os.path.exists(save_dir): | |
| os.mkdir(save_dir) | |
| checkpoint_path = os.path.join(save_dir, f"epoch_{epoch}.pth") | |
| torch.save(save_states, checkpoint_path) | |
| def save_best_checkpoint(model, model_ema, epoch, work_dir=None): | |
| save_dir = os.path.join(work_dir, "checkpoint") | |
| save_states = {"epoch": epoch, "state_dict": model.state_dict()} | |
| if model_ema != None: | |
| save_states.update({"state_dict_ema": model_ema.module.state_dict()}) | |
| if not os.path.exists(save_dir): | |
| os.mkdir(save_dir) | |
| checkpoint_path = os.path.join(save_dir, f"best.pth") | |
| torch.save(save_states, checkpoint_path) | |