This checkpoint loads fine, but it's broken
This checkpoint that I generated successfully loads via `torch.load`, but it's corrupted.
Even if PyTorch doesn't give an early signal, issues appear later with extra code, making it frustrating to debug.
Weightlens flags the exact layer and failure type in one `analyze` command.
You can try running the demo yourself, Weightlens will not fail.
Your upvotes and feedback are welcome!
Words have more power than we think. Be kind.