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
Hello,
In the pytorch implementation (https://github.com/lu-group/sbinn/blob/main/sbinn/sbinn_pytorch.py#L131), the input layer size is incorrectly defined to be #6. As per the paper only time is the input right? Also in the TF implementation the layer size is one.
I think the correct version should be net = dde.maps.FNN([1] + [128] * 3 + [6], "swish", "Glorot normal")
The text was updated successfully, but these errors were encountered:
We use an input transform turning t into 0.01*t, and periodic features of sin(t), sin(2t), etc. In total, we have 6 features which is 6 inputs. Pytorch needs to know that there are 6 inputs. Tensorflow automatically figures out that there are 6 inputs, which is why it appears that there is only 1 input in the code.
Hello,
In the pytorch implementation (https://github.com/lu-group/sbinn/blob/main/sbinn/sbinn_pytorch.py#L131), the input layer size is incorrectly defined to be #6. As per the paper only time is the input right? Also in the TF implementation the layer size is one.
I think the correct version should be
net = dde.maps.FNN([1] + [128] * 3 + [6], "swish", "Glorot normal")
The text was updated successfully, but these errors were encountered: