This sounds like something the documentation should cover, however, it doesn't and in the many tutorials I've seen nobody seems to have figured this out.
I'm trying to run onnx runtime web with a BERT model exported from hugging face. I do get all the steps working and the predictions, however I'm trying to find a built-in way to apply softmax to my predictions to get the probabilities.
From ONNX web documentation I can see the softmax operation is supported.
But from the API I have no clue on how to invoke it.
Does anyone know how to do this?
Thanks!