Joe Wallwork
Joe Wallwork
Closes #305. Brings much-needed clarity on the overloaded assignment operator. This PR is somewhat of a companion of #373, in that I discovered some of this stuff while working on...
This is now a high priority because a bug snuck in without us noticing, as reported in #365.
Closes #249. Tidies up the test suite to just run with `ctest` directly, rather than navigating into subdirectories. A big advantage of this approach is that we can use the...
Whilst working on #239, I figured out how to pass unused arguments to `ctest`. (See [249_ctest-args](https://github.com/Cambridge-ICCS/FTorch/tree/249_ctest-args).) This could be useful for, e.g., running specific tests from the `run_test_suite` script. However,...
This issue provides a list of API-breaking tasks that are to be postponed until the v2.0 release. * Make the `layout` argument to `torch_tensor_from_array` optional rather than interfacing to the...
The Torch C++ API includes a `Scalar` class, which should streamline handling scalar values in the autograd work. We should wrap this class as `torch_scalar`, as well as it's core...
Closes #122. This PR provides a development `spack.yaml` configuration file to install FTorch's dependencies. It also sets up a workflow for building a Docker image with this Spack setup inside...
Are there any code coverage tools that might work for us, accounting for both C++ and Fortran?
The segfault was resolved by ensuring `use ftorch, only : assignment(=)` was present. So that is good news. The bad news I guess is that this can be a tricksy...
This PR adds an additional CI step to build and run one of the examples in a "standalone" fashion. This checks that we can build it with CMake separately from...