F# 6.0 -> 7.0: A unique overload for method could not be determined

Please provide a succinct description of the issue.

I had the following TorchSharp code working fine in F# 6.0:

let xs = torch.zeros([|10L|], dtype=torch.float32)

After upgrading to F# 7.0 the compiler started complaining:

*A unique overload for method 'zeros' could not be determined based on type information prior to this program point. A type annotation may be needed.Known types of arguments: int64 array * dtype: torch.ScalarType * device: torch.Device * requiresGrad: bool

Candidates:
- torch.zeros(size: ReadOnlySpan<int64>, ?dtype: Nullable<torch.ScalarType>, ?device: torch.Device, ?requiresGrad: bool) : torch.Tensor 
- torch.zeros(size: ReadOnlySpan<int64>, ?dtype: Nullable<torch.ScalarType>, ?device: torch.Device, ?requiresGrad: bool) : torch.Tensor 
- torch.zeros(size: int64 array, ?dtype: Nullable<torch.ScalarType>, ?device: torch.Device, ?requiresGrad: bool) : torch.Tensor*

It does work in F# 7.0 if I change the code to:

let xs = torch.zeros([|10L|], ?dtype=Some torch.float32)

… but having to sprinkle this all over the code (in this way) is clumsy and hinders readability.

Yes, there’s a workaround,
but from a usability perspective, I’d call this a regression.

[NinoFloris]

After discussing privately with @dsyme we concluded that for a fix here we should prefer nullable TDCs. I’ve implemented it in https://github.com/dotnet/fsharp/pull/14319. I have also opened https://github.com/dotnet/fsharp/issues/14318 to track the previously mentioned example which reproduces in F# 6 of an ambiguous set that clearly shouldn’t be.

[T-Gro]

Can I mark as closed by https://github.com/dotnet/fsharp/pull/14319 now ?