'model.eval()' vs 'with torch.no_grad()'

[Naman-ntc (Naman Jain)]

Ahh with torch.no_grad() you’ll have much higher speeds and can use larger validation batch sizes so it’s useful if not recommended