interesaaat / torchsharp Goto Github PK

In Pytorch loss functions in torch.nn work as follows:

loss = NLLLoss(SOME PARAMETERS IF NEEDED)
prediction = model.forward(input)
output = loss(prediction, target)

We should match the behavior where once initialized the loss can be applied to the prediction and the target tensors.

When you right click and Run Custom Tool, it should generate the .cs file.
If it is not currently set up to do so, could can delete the .tt file and add it back to the solution.

When I try to run the template using the Run Custom Tool it gives me an error.

Could you fix it? (or maybe we don't need the template?)

Hi @interesaaat,

Thanks for the big effort to bring training to TorchSharp. Do you have a rough estimate of when all your efforts will be merged to xamarin/TorchSharp, and release on nuget?

Best regards
Mads