Jonáš Kulhánek
Jonáš Kulhánek
I have a very simple example to reproduce: ``` device = torch.device("cpu") c2 = torch.randint(0, 10, (32, 3,), dtype=torch.int32, device=device) f2 = torch.rand(size=(32, 1), dtype=torch.float32, device=device) data = sparse_collate([ SparseTensor(...
The fix is here #30.
Should I close this issue or should I link a feature request from the PyTorch repository?
Any progress here?
I agree. This is the official template and many CTU works are written using the template. Unfortunately, all these works break the license.
Inria’s impl us also correct. The ingp’s one, however, supports more degrees.
I think the issue is storing all paths in yaml files as PosixPaths instead of strings. We could perhaps implement custom yaml loader/serialization to avoid the issue? For now can...
What do you mean by LLFF dataloader? Do you mean the loader for the LLFF dataset with forward-facing scenes? The NDC scene contraction is not currently implemented.
I believe external repos should fix an NS version. The development breaks lots of stuff between versions so I think its fine. Perhaps keeping a deprecation warning for a couple...