Elias Vansteenkiste

device_id keyword changed to device, so you can also use .cuda(device=gpu_id)

You can write a wrapper dataset class and add the augmentation to a transformation class. See: torchvision.transforms.compose

Having exactly the same issue but with a machine with 2xA10, RHEL 8.6, driver 470.

Build error for snappy: ``` + cmake . -DCMAKE_INSTALL_PREFIX:PATH=/Users/elias_v/external_repositories/ffmpeg-on-apple-silicon/workdir/sw -DLIBTYPE=STATIC -DENABLE_SHARED=off -- The C compiler identification is AppleClang 14.0.3.14030022 -- The CXX compiler identification is AppleClang 14.0.3.14030022 -- Check for...

Similar issue for vid-stab. It seems to rely on SSE2 which is not available in arm. ``` Start compiling Vid-stab + cd /Users/elias_v/external_repositories/ffmpeg-on-apple-silicon/workdir/compile + cd vid.stab-1.1.0 + patch -p1 patching...

libx265 also failed but changing this helped: -DENABLE_ASSEMBLY=OFF but this will come at the cost of encoding speed

I only have an edge detector specialized for road traffic scenes.

@czczup @YeShenglong1 @zhiqi-li @Zeqiang-Lai Any update on the release of the weights for InternImage-H with DINO?