
Description of the issue
In the above code, I mistyped the word torch.nn.functional.gauss{ui}an_nll_loss and the compiler complained not that gaussuan_nll_loss is missing, but rather that whole nn is missing which is incorrect as well as misleading (I've spent lots of time trying to figure out why nn would be missing.
Sample code
#r "nuget: TorchSharp, 0.99.0"
#r "nuget: libtorch-cpu-linux-x64, 1.13.0.1"
open TorchSharp
let xs = torch.ones(10)
let loss = torch.nn.functional.gaussuian_nll_loss(xs, xs, xs)
Expected behavior
Error: input.fsx (4,18)-(4,20) typecheck error The type 'torch.nn.functional' does not define the field, constructor or member 'gaussuan_nll_loss'.
Actual behavior
Error: input.fsx (4,18)-(4,20) typecheck error The type 'torch' does not define the field, constructor or member 'nn'.
Operating system
- Ubuntu Linux 22.04, dotnet sdk 6.0.110
Same problem happens with .ipynb or .fsx files.
Description of the issue
In the above code, I mistyped the word
torch.nn.functional.gauss{ui}an_nll_lossand the compiler complained not thatgaussuan_nll_lossis missing, but rather that wholennis missing which is incorrect as well as misleading (I've spent lots of time trying to figure out whynnwould be missing.Sample code
Expected behavior
Actual behavior
Operating system
Same problem happens with
.ipynbor.fsxfiles.