when code running, calling Class FullyConnectedTensorProduct(TensorProduct), the dimension of x.shape[-1] and self._in1_dim are all 7 at the beginning, it's ok to go on. and then when code second time calling Class FullyConnectedTensorProduct(TensorProduct) , bug happen, The dimension of x.shape[-1] change into 512, and the dimension of self._in1_dim change into 514. How can I solve this problem? thank you!
when code running, calling Class FullyConnectedTensorProduct(TensorProduct), the dimension of x.shape[-1] and self._in1_dim are all 7 at the beginning, it's ok to go on. and then when code second time calling Class FullyConnectedTensorProduct(TensorProduct) , bug happen, The dimension of x.shape[-1] change into 512, and the dimension of self._in1_dim change into 514. How can I solve this problem? thank you!