pytorch報錯：ValueError: Expected more than 1 value per channel when training

pytorch報錯：ValueError: Expected more than 1 value per channel when training, got input size torch.Size([1,512,1,1]) 2.網上查找下原因爲模型中用了batchnomolization，訓練中用batch訓練的時候當前batch恰好只含一個sample，訓練中用batch訓練的時候，應該是

>>阅读原文<<