1、non-finite loss, ending training tensor(nan, device=‘cuda:0‘,2、‘LogSoftmaxBackward3、Function ‘MulB
WARNING:non-finiteloss,endingtrainingtensor(nan,device='cuda:0',grad_错误1:WARNING:non-finiteloss,endingtrainingtensor(nan,device=‘cuda:0’,grad_错误2:Function‘LogSoftmaxBackward’returnednanvaluesinits0tho