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 369 forks source link

Does Inference work on CPU alone? #289

Closed sruteesh-pivot closed 5 years ago

sruteesh-pivot commented 5 years ago

Hello , I tried running the inference using a CPU machine. But I keep getting this error

NameError: name 'build_forward_pass_graph_fn' is not defined

Does the script not support inference/training on CPU.

Thanks.

vsl9 commented 5 years ago

Which version / branch do you use? Can you please try the latest master branch? It doesn't have any build_forward_pass_graph_fn in the code.

sruteesh-pivot commented 5 years ago

Hi, I tried with use_horovod =False and I was getting shape mismatch error. I guess I was trying to use the mixed precision model on my CPU and hence the errors.

I was able to sucessfully infer using the wiki2-fp32 model.

Thanks.

ffxz commented 5 years ago

how do you use cpu to infer? @sruteesh-pivot