Closed nimz closed 4 years ago
This is minor, but the three PyTorch layers defined in the README should be put on the GPU, e.g.
norm_lstm_layer = haste.LayerNormLSTM(input_size=128, hidden_size=256, zoneout=0.1, dropout=0.05).cuda()
since the input is a CUDA tensor.
Thanks so much for pointing this out, @nimz. Fixed!
This is minor, but the three PyTorch layers defined in the README should be put on the GPU, e.g.
norm_lstm_layer = haste.LayerNormLSTM(input_size=128, hidden_size=256, zoneout=0.1, dropout=0.05).cuda()
since the input is a CUDA tensor.