Using Pytorch Frame in assigning node features

[parsisoftware]

Hello,
Can you please guide me how to use the PyTorch frame encodings (https://github.com/pyg-team/pytorch-frame/tree/master) for the node features while loading the data from csv using the following link:
https://pytorch-geometric.readthedocs.io/en/latest/tutorial/load_csv.html

Can I simply replace the encoding classes with the one defined in the above link like the following:

from torch_frame import stype
from torch_frame.nn import (
EmbeddingEncoder,
FTTransformer,
LinearEmbeddingEncoder,
LinearEncoder,
)

movie_x, movie_mapping = load_node_csv(
movie_path, index_col='movieId', encoders={
'title': LinearEmbeddingEncoder(),
'genres': EmbeddingEncoder()
})

[rusty1s]

We are working on examples and some tutorials right now. The current idea is to store TensorFrame from pytorch-frame as attribute in Data. You can then encode this TensorFrame using pytorch-frame, and pass it further down to PyG.

[denadai2]

this would be awesome, especially if we can use it in our gnn models

[rusty1s]

We have pushed an example in https://github.com/snap-stanford/relbench, see https://github.com/snap-stanford/relbench/blob/main/relbench/external/graph.py and https://github.com/snap-stanford/relbench/blob/main/relbench/external/nn.py. This might be helpful. However, we still work on a bringing an example to PyG directly.

[parsisoftware]

When the example in PyG will be ready?