[D] Are PyTorch high-level frameworks worth using?

In an attempt to better track experiment results and hyperparameters, not only did I learn about the Weights and Biases library but also ended up finding out about frameworks such as PyTorch Lightning and Ignite. I've always used raw PyTorch, so I'm not sure if these frameworks are really useful. I mostly work with academic research, right now I also need to keep track of the MAE since it's a regression problem and I don't know if these frameworks support …

biases experiment frameworks ignite learn library lightning machinelearning pytorch pytorch lightning raw results weights and biases work