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 371 forks source link

How to use TensorRT for Speech Synthesis?. #474

Open ajaysg-zz opened 4 years ago

ajaysg-zz commented 4 years ago

I have trained models of Tacotron-2 and Wavenet using the config files provided in example_configs/text2speech/.. folder. I want to reduce the inference time. Can I use TensorRT on Wavenet model for Inference?. If so how to do it?.

borisgin commented 4 years ago

You can set flags: 'use_trt': True,
See more flags here model.py

ajaysg-zz commented 4 years ago

@borisgin thank you very much. Is it enough to set the flag only in model.py? or do I have to set the flag in the config file also which I used from example_configs/text2speech/.. folder?.

borisgin commented 4 years ago

You should set flag in the config file, not in model.py.

On Thu, Jul 11, 2019 at 3:08 PM ajaysg notifications@github.com wrote:

@borisgin https://github.com/borisgin thank you very much. is it enough to set the flag only at model.py? or i have to change in the config files present in example_configs/text2speech/.. folder.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/NVIDIA/OpenSeq2Seq/issues/474?email_source=notifications&email_token=ABZRNKJOLVSPPPCTS2KZNE3P64PEVA5CNFSM4IAW56G2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODZWPXUY#issuecomment-510458835, or mute the thread https://github.com/notifications/unsubscribe-auth/ABZRNKOOBKWMJ2UYADCNIXTP64PEVANCNFSM4IAW56GQ .

ajaysg-zz commented 4 years ago

@borisgin thank you for your prompt response. I have used wavenet_float.py for training. I could not find 'use_trt' flag in that.