Aleksandr Ovechko

Taking a global strong reference to view object is a bad idea: ```swift import Foundation import ARKit private var view: Any? // this must be instance & weak variable! private...

I have similar error when model runs on GPU, my device is iPhone 8 (Apple A11 Bionic processor), iOS 12.

I'll add an android example later. That's very simple: 1) just try android projects with **statically linked** Tensorflow NDK library: https://github.com/search?q=filename%3AAndroid.mk+static+tensorflow.a 2) and add TFSecured dir (source files) to `Android.mk`...

@Volutionn I implemented the `deform_conv2d` operation using CoreML custom layers: https://github.com/dneprDroid/DeformConv2d-Metal It's GPU-accelerated and supports both iOS and macOS. You can try the demo app and the example converter script...

@zack-dev-cm I've implemented a CoreML custom layer and a MIL op for `torch.linalg.lstsq`. It supports multi-batch input tensors and [`gelsd`](https://pytorch.org/docs/stable/generated/torch.linalg.lstsq.html#torch.linalg.lstsq) driver: https://github.com/dneprDroid/Lstsq-CoreML You can try the demo app and the...

@SpiraMira please try this: ```python import torch from torch import nn import coremltools as ct def my_cdist(x1: torch.Tensor, x2: torch.Tensor, p=2.0): assert p == 2.0 x1_norm = x1.pow(2).sum(-1, keepdim=True) x1_pad...