-
### 🐛 Describe the bug
We found there are ~180 cases got failed specifically on MTL, it can pass on PVC, on ARC the cases will fail in op creation with fp64 cases excluded.
MTL:
[mtl.log](https:/…
-
redesign MTD's convolution code to handle the presence of bad data.
-
### 🚀 The feature, motivation and pitch
Recurrent LLM architectures such as Mamba (v1 and v2) and xLSTM use depthwise convolution and GroupNorm in each block. Both are currently not supported by Pyto…
-
The 2D differential convolution method seems to be equivalent to using Lagrange polynomial padding of degree kernel size - 1. Here is a minimal example supporting kernel sizes 3, 5, 7, using vertical …
-
For low latency, you can get better loss tolerance for a given delay from a convolutional code such as https://github.com/catid/CauchyCaterpillar ... You may want to consider using this FEC (or one li…
-
Hi,
We are having issues with our categorizer training stopping after only ~10-13 iterations. We are seeing warnings like:
"Found untraced functions such as _jit_compiled_convolution_op, _jit_c…
-
### Search before asking
- [X] I have searched the Ultralytics YOLO [issues](https://github.com/ultralytics/ultralytics/issues) and found no similar bug report.
### Ultralytics YOLO Component
Val
…
-
I would like add animations for [depth wise](https://animatedai.github.io/media/depthwise-convolution-animation-3x3-kernel.gif) and [point wise](https://animatedai.github.io/media/depthwise-separable-…
-
https://github.com/huang-yh/GaussianFormer/blob/3bf28d50b41c8a54fda8147defe52a08f1a05b4e/model/encoder/gaussian_encoder/spconv3d_module.py#L53-L67
Although interaction can be efficient using spar…
-
It would be nice to have FFT based convolution supported in mlx. FFT bases convolution shows much better performance for large images / arrays and kernels. The FFT building blocks are already support…