1ytic / warp-rnnt

CUDA-Warp RNN-Transducer
MIT License
211 stars 41 forks source link

Not support pytorch 1.6 #7

Closed wyattsuen closed 3 years ago

wyattsuen commented 3 years ago

Hi, thanks for this package. But the warp-rnnt can't support pytorch 1.6. Would you kindly update a newer version which support the newest pytorch and a cpu version?

1ytic commented 3 years ago

In order to install a custom gpu torch extension, you need to install CUDA tools and libraries independently from python cudatoolkit package. Make sure that you have installed CUDA in /usr/local/cuda. I tested with CUDA 11.1 and PyTorch 1.6 and cudatoolkit 10.2. All test works fine: python -m warp_rnnt.test.