Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

pre_transform works not as expected

See original GitHub issue

I want to run an expensive transform to modify data on vertices, so I think I should pass it to pre_transform key – it would run a transform on every graph in the dataset and will cache processed dataset.

But it works not as expected. For example, take OneHotDegree transform. Note that I delete dataset folder on every run of the following snippets:

No transform

dataset = TUDataset(root="../data/", name="MUTAG")
dataset[0].x

tensor([[1., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0.],
        [0., 1., 0., 0., 0., 0., 0.],
        [0., 0., 1., 0., 0., 0., 0.],
        [0., 0., 1., 0., 0., 0., 0.]])

Transform

dataset = TUDataset(root="../data/", name="MUTAG", transform=OneHotDegree(5))
dataset[0].x

tensor([[1., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0., 0.],
        [0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],
        [0., 0., 1., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],
        [0., 0., 1., 0., 0., 0., 0., 0., 1., 0., 0., 0., 0.]])

Pre-transform

dataset = TUDataset(root="../data/", name="MUTAG", pre_transform=OneHotDegree(5))
dataset[0].x

tensor([[0., 0., 1., 0., 0., 0.],
        [0., 0., 1., 0., 0., 0.],
        [0., 0., 1., 0., 0., 0.],
        [0., 0., 0., 1., 0., 0.],
        [0., 0., 0., 1., 0., 0.],
        [0., 0., 1., 0., 0., 0.],
        [0., 0., 1., 0., 0., 0.],
        [0., 0., 1., 0., 0., 0.],
        [0., 0., 0., 1., 0., 0.],
        [0., 0., 0., 1., 0., 0.],
        [0., 0., 1., 0., 0., 0.],
        [0., 0., 1., 0., 0., 0.],
        [0., 0., 0., 1., 0., 0.],
        [0., 0., 1., 0., 0., 0.],
        [0., 0., 0., 1., 0., 0.],
        [0., 1., 0., 0., 0., 0.],
        [0., 1., 0., 0., 0., 0.]])

I expect to have the result of transform (case 2) in the case of pre-transform too. Is it bug or feature?

How should I apply expensive transforms and get the data modified as in the case 2?

Issue Analytics

State:
Created a year ago
Comments:5 (1 by maintainers)

Top GitHub Comments

1reaction

rusty1scommented, May 17, 2022

This will be fixed in https://github.com/pyg-team/pytorch_geometric/pull/4669. It was caused by the weird interplay of TUDataset and the detection of “categorical” features induced by the use_node_attr argument.

0reactions

oleg-kachancommented, May 17, 2022

Moreover, as a pre-transform it accepts only a vector of all ones, truncating the matrix of all ones to a single column

class ReplaceBlock(BaseTransform):

    def __init__(self, value, width=2):
        self.value = value
        self.width = width
    
    def __call__(self, data):
        data.x = torch.tensor(self.value).reshape(-1, 1).repeat(data.num_nodes, self.width)
        return data

dataset = TUDataset(root="../data/", name="MUTAG", transform=ReplaceBlock(1.0, width=2))
dataset[0].x

tensor([[1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.],
        [1., 1.]])

dataset = TUDataset(root="../data/", name="MUTAG", pre_transform=ReplaceBlock(1.0, width=2))
dataset[0].x

tensor([[1.],
        [1.],
        [1.],
        [1.],
        [1.],
        [1.],
        [1.],
        [1.],
        [1.],
        [1.],
        [1.],
        [1.],
        [1.],
        [1.],
        [1.],
        [1.],
        [1.]])

Top Results From Across the Web

HDK: Object Transform Model

Use this function to convert co-ordinates that have not been transformed by the follow path or pre-transform. Typically, this function is only used...

Custom function transformer not performing as expected

The transformer seems to work on it's own, and the fit() and transform() methods work individually, but when I include it in a...

API Changes — Matplotlib 1.5.1 documentation

It is not expected that this will affect any users as the current ids are generated from an md5 hash of properties of...

Pretransformation Rules

Use a pretransformation rule to populate an order attribute or set its default value before Order Management transforms the source order.

Solved: Custom Transform Node & xform - Maya

Using the joint's extra pre-transform attributes works pretty good, except for non-uniform scaling because of how inverseScale works.