The current cudnn lstm only takes h, c and params as input. The layer norm is not availiable. Is there a way to add layer norm or when is the layer norm gonna be available?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| LSTMP doesn't work in cudnn 7.1.4 | 2 | 3180 | February 18, 2021 | |
| LSTM layer do not support peephole! | 0 | 488 | August 24, 2018 | |
| Feature request : Cudnn Layer Normalization | 1 | 1458 | January 14, 2020 | |
| Batch normalization implementation using cuDNN | 1 | 2103 | November 20, 2020 | |
| inconsistency in cudnnRNNMode_t | 0 | 663 | May 3, 2016 | |
| Ford Pass using weights of tf.compat.v1.keras.layers.CuDNNLSTM | 4 | 3122 | June 18, 2020 | |
| Use of cuDNN RNN | 16 | 6875 | September 20, 2018 | |
| How to implement a complete LSTM network with cudnn? | 5 | 3780 | January 13, 2020 | |
| API Reference :: NVIDIA Deep Learning cuDNN Documentation | 1 | 798 | January 4, 2021 | |
| Optimizing Recurrent Neural Networks in cuDNN 5 | 17 | 790 | August 8, 2018 |