Graphbook should have a utility function that inspects a given function's name, signature, and docs and returns a Step class with those attributes. The input should be the first parameter. We should construct the Parameters based on the function signature starting from the second parameter and each of its annotations. i.e:
- str -> "string"
- int -> "number"
- float -> "number"
- tuple -> "list[number]"
- etc.
This function should inspect pytorch/numpy style functions that typically have the Tensor or ndarray as the first parameter. It should work for functions like torch.mean, torch.std and numpy.mean, and functions from the package torchvision.transforms.functional.
Graphbook should have a utility function that inspects a given function's name, signature, and docs and returns a Step class with those attributes. The input should be the first parameter. We should construct the Parameters based on the function signature starting from the second parameter and each of its annotations. i.e:
This function should inspect pytorch/numpy style functions that typically have the Tensor or ndarray as the first parameter. It should work for functions like torch.mean, torch.std and numpy.mean, and functions from the package torchvision.transforms.functional.