I upgraded to tensorflow 2.5.0 in new anaconda environment; Cuda 11.2 GPU Geforce 1060 (Windows laptop); numpy default 1.20.3
I am having an error in running RNN/GRU/LSTM codes that have been working fine in previous versions. No issues on CNN (1D/2D) and DNN (Dense layers).
The error msg: “NotImplementedError: Cannot convert a symbolic Tensor (lstm_2/strided_slice:0) to a numpy array. This error may indicate that you’re trying to pass a Tensor to a NumPy call, which is not supported”
Looking for help: Any suggestion from anyone?
Problem is with numpy version.
Tensorflow 2.5.0 version requires Numpy 1.19.5
!pip install tensorflow
Default it will install numpy 1.19.5
@Ashok_Bhowmick, Can you please try using numpy==1.19.5 and let us know? Thanks!