Francesco Ceccon
Francesco Ceccon
Yep, I can make a PR for this.
The description for `write_input_bounds` looks good. I would change the description of `load_input_bounds` to: ``` Read the input bounds from the given file. The input bounds file should be generated...
At the moment with ONNX we can support keras, tensorflow, and pytorch. Scikit has an ONNX exporter but I'm not sure how it works with neural networks.
Hey, yes I think we should have it. Argent added `starknet.isPreauthorized()` for this use-case. I like the wagmi api for this where you pass a `autoConnect` prop to the `StarknetProvider`...
I cannot reproduce exactly your issue, but it's true that with a production build auto connect doesn't work. Will fix it.
Even with `setTimeout(..., 0)` the issue persist. Adding a 100ms delay fixes this issue but it's kinda horrible.
That does solve the issue. It feels like the argent detection code could be improved, but I don't know exactly how.
Interesting, it looks like your issue is that it takes much longer to inject `starknet` in the window. This does look odd and I can't reproduce it. Can you try...
It looks odd indeed. Let's ship the workaround that fixes the issue sometimes but let's keep this issue open.
The type definition in starknet.js is wrong. The return value is a weird array with keys(!), the objects are deserialized from a flat felt array to their type (so for...