search

nitishsrivastava / deepnet

Implementation of some deep learning algorithms.
BSD 3-Clause "New" or "Revised" License
893 stars 438 forks source link

CUDAMatException: Incompatible matrix dimensions #102

Open TianCuan opened 7 years ago

TianCuan commented 7 years ago

I am a beginner of deepnet and I met this problem:

Train Step: 0Traceback (most recent call last): File "/home/deepnet/trainer.py", line 60, in main() File "/home/deepnet/trainer.py", line 54, in main model.Train() File "/home/deepnet/neuralnet.py", line 649, in Train losses = self.TrainOneBatch(step) File "/home/deepnet/dbm.py", line 270, in TrainOneBatch losses1 = self.PositivePhase(train=True, step=step) File "/home/deepnet/dbm.py", line 133, in PositivePhase self.ComputeUp(node, train=train) File "/home/deepnet/dbm.py", line 80, in ComputeUp layer.GetData() File "/home/deepnet/replicated_softmax_layer.py", line 91, in GetData self.state.assign(self.data) File "/home/deepnet/cudamat/cudamat.py", line 340, in assign raise generate_exception(err_code) cudamat.cudamat.CUDAMatException: Incompatible matrix dimensions.

how can i resolve it ?