Joe Wallwork

Results 45 issues of Joe Wallwork

I worked through the tutorials using Cylc 8.0.2. When I tried to run the part of the tutorial that uses `cylc review` I got the error message "cylc review has...

Closes #115. Turns existing examples into integration tests using CTest.

The ability to do online training in FTorch would be really useful. Presumably achieving this would mainly involve wrapping other parts of the (Py)Torch API. @TomMelt suggested that this piece...

enhancement
autograd

Hi, thanks very much for your work on `pytest-mpi` - we've found it really useful! I think we might have come across a bug where Pytest hangs upon reaching an...

When we run `examples/1_SimpleNet/simplenet.py`, the final thing that's executed is effectively ``` a = [0.0, 1.0, 2.0, 3.0, 4.0] model(torch.Tensor(a)) ``` This would also work with batching e.g., ``` a...

enhancement
good first issue
hackathon

Closes #138.

enhancement

Closes #151. Implements `torch_to_blob` on the C++ side and `torch_tensor_to_array` on the Fortran side. Updates example 1 to use this.

enhancement

Related to #111. Recently we've discussed implementing automatic differentiation (AD) in FTorch using the underlying autograd AD tool built into Torch. The first step to achieve this would be to...

enhancement
online-training

FTorch implements `torch_tensor_from_array` to map Fortran arrays to (F)Torch tensors. It would be useful to have the reverse mapping, too. This would provide a way to get around the issue...

enhancement
online-training

Related to #111. This will be needed to complete the autograd example in #139. Probably best to do this after that PR is merged, otherwise it will become unwieldy.

enhancement
online-training