Tensorflow-Directml-Plugin With Cuda Commands

[Om1f]

Hello everyone. I'm new in this field. I'm working on some nns. However, when I'm using NN like LSTM, the compailer returns error for CUDnn commands to there aren't any commands. I'm using AMD RX7600. Additionally, torch-directml has a problem like this. I'm curcious about when will us use all NN like LSTM or GRU?

[Om1f]

Because I'm regreting to bought AMD GPU and I must use these NN with my CPU and its so slow.