This program of training as if can only be done once and again training data and the last time I'd like to call the latest checkpoint continue training always go wrong,
Would like to ask how to resuming training, his checkpoint save is certainly useful, finally also said to strengthen training cycle, can you tell me how to achieve
CodePudding user response:
An error is the ValueError: Tensor 's shape (1174, 128) is not compatible with supplied shape (1138, 128)