(args, dl, model, optimizer, loss_func, device)
| 324 | return poses |
| 325 | |
| 326 | def eval_on_epoch(args, dl, model, optimizer, loss_func, device): |
| 327 | model.eval() |
| 328 | val_loss_epoch = [] |
| 329 | for data, pose in dl: |
| 330 | inputs = data.to(device) |
| 331 | labels = pose.to(device) |
| 332 | if args.preprocess_ImgNet: |
| 333 | inputs = preprocess_data(inputs, device) |
| 334 | predict = model(inputs) |
| 335 | loss = loss_func(predict, labels) |
| 336 | val_loss_epoch.append(loss.item()) |
| 337 | val_loss_epoch_mean = np.mean(val_loss_epoch) |
| 338 | return val_loss_epoch_mean |
| 339 | |
| 340 | |
| 341 | def train_on_epoch(args, dl, model, optimizer, loss_func, device): |
no test coverage detected