Xenova/mobilevit-small · Need to modify the pipeline code snippet sample

It seems that Xenova/mobilevit-small model should be loaded with { quantized: false }. When I set the parameter to { quantized: true }, the model produces false predictions. I cannot explain why this behavior happens, but this peculiarity needs to be noted in the API documentation, e.g Use in Transformer.js pop-up window. The blog article posted by the user "Xenova" is in agreement with my finding that { quantized: true } should be applied. See the "Running the model in the browser" section of https://github.com/huggingface/blog/blob/main/ml-web-games.md.

I would also be thankful if someone explains why this behavior happens. Thanks.