Carlos Mocholí
Carlos Mocholí
I believe #18105 will help here
We can link to the PyTorch docs so that the user can easily see the available options for the selected torch version. I don't want us to list them in...
PyTorch is also considering supporting fsspec paths automatically: https://github.com/pytorch/pytorch/issues/118036
@Borda do not lock the issue so that the community can request being added/removed.
Probably yes, if GitHub triggers API events for discussions too https://github.com/carmocca/probot/blob/87ca7b14202ea6afce291e5ee5b9a6c40c2a99df/src/auto-cc-bot.ts#L77-L79
The next torchmetrics release should avoid this problem thanks to https://github.com/Lightning-AI/torchmetrics/pull/2316
The jsonargparse-y way of doing this would be to instead specify which Optimizer class you want to select to let the parser pull out the arguments of said class. For...
Yes, we cannot have jsonargparse instantiate the class directly for that reason. But you can still tell it to add all the arguments of a class (or classes) into a...
You can start by understanding this minimal example: ```python import torch import jsonargparse parser = jsonargparse.ArgumentParser() parser.add_subclass_arguments(torch.optim.Optimizer, "optimizer", instantiate=False, fail_untyped=False, skip={"params"}) args = parser.parse_args() print(args) ``` ```shell python example.py --optimizer...
And here's how you would use the above to instantiate the optimizer: ```python from typing import Any, Tuple, Dict, Union def instantiate_class(args: Union[Any, Tuple[Any, ...]], init: Dict[str, Any]) -> Any:...