search

NVIDIA / flowtron

Flowtron is an auto-regressive flow-based generative network for text to speech synthesis with control over speech variation and style transfer
https://nv-adlr.github.io/Flowtron
Apache License 2.0
887 stars 177 forks source link

Accelerated inference with TensorRT #148

Open Jcwscience opened 2 years ago

Jcwscience commented 2 years ago

I know on the FastSpeech GitHub project there is a documented way to convert the speech model through ONNX to run with TensorRT. Has anyone attempted this with Flowtron? I’m using an AGX Xavier and I figure there could be a performance increase.

farzanehnakhaee70 commented 2 years ago

Hi, This link represents a script for converting the model to ONNX.

However, the performance was worse than the pure pytorch model.

farzanehnakhaee70 commented 2 years ago

Is there any one has any improvement for the model acceleration in TRT or ORT?