[FEATURE] Handle cliptokenizer.onnx in Stable Diffusion sample
Describe the bug There is no way to get the cliptokenzier.onnx that is used by the Stable Diffusion sample other than exporting the entire project
@nmetulev would having an implementation in Microosft.ML.Tokenizers help here?
100%. Right now, we depend on an cliptokenizer.onnx model which complicates the implementation.
Great! We're tracking that here.
dotnet/machinelearning#6993
In the meantime though, if you wanted a C# implementation, this might work for your scenario.
https://github.com/LittleLittleCloud/Torchsharp-stable-diffusion-2/blob/main/Tokenizer.cs
cc: @LittleLittleCloud to confirm this is all that's needed for tokenization of Stable Diffusion.