Maybe for a later release, we could think about adding pytorch and tensorflow. The DS-1000 benchmark uses them so, 1) we should not copy test-cases from there. But 2) if we find missing things, that we want to cover, these dependencies might be useful.
Maybe for a later release, we could think about adding pytorch and tensorflow. The DS-1000 benchmark uses them so, 1) we should not copy test-cases from there. But 2) if we find missing things, that we want to cover, these dependencies might be useful.