You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I feel we should follow Pytorch to add equations to APIs --- for example, https://pytorch.org/docs/stable/nn.html#torch.nn.Conv1d. I personally believe that is how a math library's documentation should look like. Also, debugging ML models is quite different from debugging typical code --- debugging ML model (such as figuring why it doesn't work) requires knowledge of the underlying mathematical models.