imgs, labels = imgs.to(device), labels.to(device)
out = net(imgs)
loss = loss_fn(out, labels)
ctr += 1
pred = out.data.max(1)[1]
running_metrics_val.update(pred.cpu().numpy(),
labels.cpu().numpy())
val_loss += loss.item()
score, class_iou = running_metrics_val.get_scores()
for k, v in score.items():
print(k, v)
# for k, v in class_iou.items():
# logger.info('{}: {}'.format(k, v))
running_metrics_val.reset()
print("EPOCH %d: VAL loss=%f" % (EPOCHS, val_loss / (ctr)))
scheduler.step(val_loss)
if score["Mean IoU : \t"] >= best_iou:
best_iou = score["Mean IoU : \t"]
torch.save(net.state_dict(), 'best_val.wts')
# if val_loss < best_val_loss:
# torch.save(net.state_dict(), 'best_val.wts')
torch.save(net.state_dict(), 'trained.wts')
running_metrics_val.reset()
val_loss_meter.reset()
#print("EPOCH %d: VAL loss=%f"%(EPOCHS, val_loss/(ctr)))
# scheduler.step(val_loss)
# if score["Mean IoU : \t"] >= best_iou:
# best_iou = score["Mean IoU : \t"]
# torch.save(net.state_dict(), 'best_val.wts')
if score["Mean IoU : \t"] >= best_iou:
best_iou = score["Mean IoU : \t"]
state = {
"epoch": EPOCHS + 1,
"model_state": net.state_dict(),
"optimizer_state": optimizer.state_dict(),
"scheduler_state": scheduler.state_dict(),
"best_iou": best_iou,
}
save_path = os.path.join(writer.file_writer.get_logdir(),
"best_val_model.pkl")
torch.save(state, save_path)
# if val_loss < best_val_loss:
# torch.save(net.state_dict(), 'best_val.wts')
#torch.save(net.state_dict(), 'trained.wts')