Soumith Chintala

@kinoc i really dont think training on TX1 is a good idea, 4GB memory in total between GPU and CPU. Either ways, loadcaffe seems to not be supporting 32-bit ARM,...

@kinoc ok in that case i can figure out the rest of the issues with TX1, but not until a week. My TX1 is at the office, and I leave...

ouch that's painful indeed. I can fix up the torch image loader in the next 7 days (ironically the COCO deadline is tonight). We already have the torch loader do...

just an infrastructure thing

I haven't fully checked whether this particular tensor layout is supported in cudnn, but cudnn supports non-contiguous tensors of certain kinds (for input / output). If I understand this correctly,...

@c-hofer look at https://github.com/zdevito/ATen/blob/31d00ab7fdf00c258b0fad5b1b05af77e92b64a9/aten/src/ATen/test/dlconvertor_test.cpp You can use the DLPack format which is a cross-framework, well-specified and simple format that we support importing from: https://github.com/dmlc/dlpack/

you probably should do: `git submodule update --init --recursive`

fyi SyncBatchNorm added now in PyTorch master via https://github.com/pytorch/pytorch/pull/14267 For documentation, see: https://pytorch.org/docs/master/nn.html#torch.nn.SyncBatchNorm

that is awesome, thank you!

Hi @marek-simonik I am running this on an iPhone 12 mini, which doesn't have LiDAR. So, it does report xyz to always be 0, 0, 0. I am wondering why...