Two devices bug.

[seblessa]

I kept getting this error:

RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking argument for argument boundaries in method wrapper_CUDA_Tensor_bucketize)

So I modified the source code, and after a few errors like the above I got it to work.

[lostella]

@seblessa could you share a minimal reproducible example for the issue?

I think you will run into the above issue if the context data you pass to chronos is allocated to the GPU. Is that what you're doing? If so, just ensuring that the input batches you pass to the model live on CPU should make it work