These changes allow for training to be completed using more than 1 GPU. Using multiple GPUs is optional, defined by a new boolean argument. This was to address a few Issues raised discussing/requesting multiple GPU.
A couple chunks of code got moved around to implement this without having to copy code in multiple spaces. I also added tracking training time.
I successfully test ran these changes on Ubuntu 18.04 LTS (Bionic Beaver) using 2 GPU.
If you like the idea and want anything changed in the pull request feel free to suggest.
These changes allow for training to be completed using more than 1 GPU. Using multiple GPUs is optional, defined by a new boolean argument. This was to address a few Issues raised discussing/requesting multiple GPU.
A couple chunks of code got moved around to implement this without having to copy code in multiple spaces. I also added tracking training time.
I successfully test ran these changes on Ubuntu 18.04 LTS (Bionic Beaver) using 2 GPU.
If you like the idea and want anything changed in the pull request feel free to suggest.