Inconsistant Results

1

I have something super weird going on. I have two networks that are initialized with the same values but use different parameters. When I do a forward pass using the same data the networks have different results!

You can see this in this jupyter notebook https://github.com/blester125/WeirdNotebook/blob/master/PyTorchUnstable.ipynb

Does anyone know what is going on?

2
  1. You got the weight dimension flipped - It must (out, in) -> (200, 784) and not (784, 200). The code never used your weights to do forward. And…
  2. You got the parameter name wrong. It must be weight and not weights
1 Like