Hi everyone,
I’m excited to share that I’m working on a new library to simplify PyTorch model training and validation by eliminating the need for long, repetitive training loops. Tools like Lightning and Ignite are awesome, but they can sometimes feel too opinionated, overly complex, or not flexible enough for certain use cases. What’s your take? My goal is to create something that’s simple yet customizable, without unnecessary abstractions. I’d love to hear your thoughts. What frustrates you about current approaches? Are there features you’ve always wanted in a training library? Let me know what you think!
Thanks!