'params': decayed_params, 'weight_decay': cfg.weight_decay }, { 'params': no_decayed_params }, { 'order_params': net.trainable_params() }] optimizer = RMSProp(group_params, lr, decay=0.9, weight_decay=cfg.weight_decay, momentum=cfg.momentum, epsilon=cfg.opt_eps, loss_scale=cfg.loss_scale) eval_metrics = { 'Loss': nn.Loss(), 'Top1-Acc': nn.Top1CategoricalAccuracy(), 'Top5-Acc': nn.Top5CategoricalAccuracy() } if args_opt.resume: ckpt = load_checkpoint(args_opt.resume) load_param_into_net(net, ckpt) model = Model(net, loss_fn=loss, optimizer=optimizer, metrics={'acc'}) print("============== Starting Training ==============") loss_cb = LossMonitor(per_print_times=batches_per_epoch) time_cb = TimeMonitor(data_size=batches_per_epoch) callbacks = [loss_cb, time_cb] config_ck = CheckpointConfig(save_checkpoint_steps=batches_per_epoch, keep_checkpoint_max=cfg.keep_checkpoint_max)
# optimizer decayed_params = [] no_decayed_params = [] for param in net.trainable_params(): if 'beta' not in param.name and 'gamma' not in param.name and 'bias' not in param.name: decayed_params.append(param) else: no_decayed_params.append(param) group_params = [{'params': decayed_params, 'weight_decay': cfg.weight_decay}, {'params': no_decayed_params}, {'order_params': net.trainable_params()}] optimizer = Momentum(params=net.trainable_params(), learning_rate=Tensor(lr), momentum=cfg.momentum, weight_decay=cfg.weight_decay) eval_metrics = {'Loss': nn.Loss(), 'Top1-Acc': nn.Top1CategoricalAccuracy(), 'Top5-Acc': nn.Top5CategoricalAccuracy()} if args_opt.resume: ckpt = load_checkpoint(args_opt.resume) load_param_into_net(net, ckpt) model = Model(net, loss_fn=loss, optimizer=optimizer, metrics={'acc'}) print("============== Starting Training ==============") loss_cb = LossMonitor(per_print_times=batches_per_epoch) time_cb = TimeMonitor(data_size=batches_per_epoch) callbacks = [loss_cb, time_cb] config_ck = CheckpointConfig(save_checkpoint_steps=batches_per_epoch, keep_checkpoint_max=cfg.keep_checkpoint_max) ckpoint_cb = ModelCheckpoint(prefix=f"shufflenet-rank{cfg.rank}", directory=cfg.ckpt_path, config=config_ck) if args_opt.is_distributed & cfg.is_save_on_master: