应使用“torch.device”或传递字符串作为参数来设置“device”参数

Question

我的数据迭代器当前在 CPU 上运行，因为device=0参数已被弃用。但我需要它与模型的其余部分等一起在 GPU 上运行。

这是我的代码：

pad_idx = TGT.vocab.stoi["<blank>"]
model = make_model(len(SRC.vocab), len(TGT.vocab), N=6)
model = model.to(device)
criterion = LabelSmoothing(size=len(TGT.vocab), padding_idx=pad_idx, smoothing=0.1)
criterion = criterion.to(device)
BATCH_SIZE = 12000
train_iter = MyIterator(train, device, batch_size=BATCH_SIZE,
                        repeat=False, sort_key=lambda x: (len(x.src), len(x.trg)),
                        batch_size_fn=batch_size_fn, train=True)
valid_iter = MyIterator(val, device, batch_size=BATCH_SIZE,
                        repeat=False, sort_key=lambda x: (len(x.src), len(x.trg)),
                        batch_size_fn=batch_size_fn, train=False)
#model_par = nn.DataParallel(model, device_ids=devices)

上面的代码给出了这个错误：

The `device` argument should be set by using `torch.device` or passing a string as an argument. This behavior will be deprecated soon and currently defaults to cpu.
The `device` argument should be set by using `torch.device` or passing a string as an argument. This behavior will be deprecated soon and currently defaults to cpu.

我尝试'cuda'作为参数传入device=0，但收到此错误：

<ipython-input-50-da3b1f7ed907> in <module>()
    10     train_iter = MyIterator(train, 'cuda', batch_size=BATCH_SIZE,
    11                             repeat=False, sort_key=lambda x: (len(x.src), len(x.trg)),
---> 12                             batch_size_fn=batch_size_fn, train=True)
    13     valid_iter = MyIterator(val, 'cuda', batch_size=BATCH_SIZE,
    14                             repeat=False, sort_key=lambda x: (len(x.src), len(x.trg)),

TypeError: __init__() got multiple values for argument 'batch_size'

我也尝试过device作为参数传入。设备被定义为device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')

但收到与上面相同的错误。

任何建议将不胜感激，谢谢。

Answer 1

pad_idx = TGT.vocab.stoi["<blank>"]
model = make_model(len(SRC.vocab), len(TGT.vocab), N=6)
model = model.to(device)
criterion = LabelSmoothing(size=len(TGT.vocab), padding_idx=pad_idx, smoothing=0.1)
criterion = criterion.to(device)
BATCH_SIZE = 12000
train_iter = MyIterator(train, batch_size=BATCH_SIZE, device = torch.device('cuda'),
                        repeat=False, sort_key=lambda x: (len(x.src), len(x.trg)),
                        batch_size_fn=batch_size_fn, train=True)
valid_iter = MyIterator(val, batch_size=BATCH_SIZE, device = torch.device('cuda'),
                        repeat=False, sort_key=lambda x: (len(x.src), len(x.trg)),
                        batch_size_fn=batch_size_fn, train=False)

经过大量的试验和错误，我设法设置device为 device = torch.device('cuda')而不是device=0