Load Tensorflow model from stream

### System information

- **OS version/distro**: Windows 10 Enterprise 64 bits
- **.NET Version (eg., dotnet --info)**: .NET Core 3.1

### Question

Is there a way to load a Tensorflow model from a memory stream instead of a filepath? Currently the only way I managed to load a TF model is by calling:
`public static TensorFlowModel LoadTensorFlowModel(this ModelOperationsCatalog catalog, string modelLocation);`

However, if I want to load a ML.NET trained model, I can do it by calling:
`public ITransformer Load(string filePath, out DataViewSchema inputSchema);`
or
`public ITransformer Load(Stream stream, out DataViewSchema inputSchema);`

I am looking for a method like the latter one but for TF models. Does anybody know if something like that is already implemented? How could I load a TF model without the need of having the .pb file in disk?
Will such a method be implemented in a future release?

Thanks for your help,
Carlos.