The codebase is mainly built with following libraries:
-
Python 3.8 or higher
-
PyTorch and torchvision.
We can successfully reproduce the main results under this setting:
Tesla A100 (40G): CUDA 11.1 + PyTorch 1.8.0 + torchvision 0.9.0 -
deepspeed==0.5.8DS_BUILD_OPS=1 pip install deepspeed -
PyAV
pip install av
- We recommend you to use
PyTorch >= 1.8.0. - We observed accidental interrupt in the last epoch when enabled
deepspeed. Therefore, we disable the option ofdeepspeedin the released version.