I’ve been beginning to use PyTorch Lightning as a layer further abstracted from PyTorch. I’ve found it to be incredibly useful, simplifying my approach to many applications without losing the flexibility that I need. It also includes some common features, such as regularizations, built-in in a way that drastically helps with training. I also found it was able to better leverage my GPU for some tasks. Has anyone else worked with Lightning? What are your thoughts? Thanks!
I have not worked with lightning or pytorch at all. I intend to learn some PyTorch for a professional development project, but presently I only have experience in TensorFlow. TensorFlow is similar to pytorch.
In TensorFlow, I started with keras, which sounds similar to lightning because it is abstracted from tensorflow. I found it to be a great entry point. I haven’t gotten too far past it yet.
It’s interesting to hear about your experience.