search

NVIDIA / OpenSeq2Seq

Toolkit for efficient experimentation with Speech Recognition, Text2Speech and NLP
https://nvidia.github.io/OpenSeq2Seq
Apache License 2.0
1.54k stars 372 forks source link

No OpKernel was registered to support Op 'CudnnRNNCanonicalToParams' #522

Open kirill-dev-pro opened 4 years ago

kirill-dev-pro commented 4 years ago

Im trying to follow steps from https://nvidia.github.io/OpenSeq2Seq/html/installation.html and https://nvidia.github.io/OpenSeq2Seq/html/speech-recognition.html#getting-started to get pretrained model working but got following error

No OpKernel was registered to support Op 'CudnnRNNCanonicalToParams' used by node ForwardPass/ds2_encoder/cudnn_gru/cudnn_gru/CudnnRNNCanonicalToParams (defined at /home/parallels/Work/OpenSeq2Seq/open_seq2seq/encoders/ds2_encoder.py:327) with these attrs: [rnn_mode="gru", seed2=252, seed=0, dropout=0, num_params=12, T=DT_FLOAT, input_mode="linear_input", direction="bidirectional"]
Registered devices: [CPU, XLA_CPU]
Registered kernels:
  <no registered kernels>

     [[ForwardPass/ds2_encoder/cudnn_gru/cudnn_gru/CudnnRNNCanonicalToParams]]

I have no idea. How could I fix this?

OndrejTexler commented 3 years ago

I am getting the same error? @Zmeu213, were you able to resolve it?

OndrejTexler commented 3 years ago

Hmm ... so it seems that installing correct version of CUDA, which is 10.0, solves this issue. Remember to update LD_LIBRARY_PATH

kirill-dev-pro commented 3 years ago

Oh yea... I think I managed to go through this but I dont remember how 😅 Sorry

kirill-dev-pro commented 3 years ago

I think my problem was on hardware side. On my macbook it failed but on pc with nvidia graphic card it works ok