Closed lixiangparadise closed 5 years ago
Because the default way to use this project is to use Docker, if you build the docker image yourself, it will install tensorflow and pytorch. You should be able to skip the docker build command now, and instead just use bmccann/decanlp:cuda9torch041 for the code on the master branch. If you don't have CUDA, then you can simply remove the cuda9 from the docker image tag.
I have installed pytorch and tensorflow in python However,when I run the comman 'docker build -t decanlp . && cd -', it would install the pytorch and tensorflow. Because of the network speed, it would be interuppted. I don't know why I have installed the pytorch and tensorflow but it still needs to download and install it.
Thank you for your answer.