cospectrum

Results 91 comments of cospectrum

> Yes, `warp_into` will guarantee that “Output pixels whose pre-image lies outside the input image are set to `default`.” So there is no need to create the image with any...

> Tried running the inference model and it runs very slow. The python code runs much faster. Can you please check what is causing this? > > Weights: https://drive.google.com/file/d/1eEhrck8zzv5HP3vUf7wiimOFk_y5Gwn0/view?usp=sharing >...

Looks like it, but I encountered a problem with grpc `traces` instead of `metrics`. Maybe there is a problem with `_logs` too.

`sfv = "0"` dependency breaks the build

> I compared onnx with PPOCR on CPU and found that using python was 9 times faster than tract. I wonder if this is normal `onnx` uses multiple threads by...

I inserted my own `mul`/`add` functions, which solved my problem: ```rust const F32_BOUND: f32 = 1e3; fn add_f32(a: f32, b: f32) -> f32 { #[cfg(kani)] { kani::assume(a.abs() < F32_BOUND); kani::assume(b.abs()...

@zhassan-aws I saw this flag. But I guess it also disables overflow checking for other types (like i32), which I don't want.

I don’t like boolean argument. We should either have 2 functions `speckle_noise` and `speckle_noise_per_channel` or keep only one of them as `speckle_noise` (without bool arg).