andreasbinder / Point-GNN-PyTorch

MIT License
6 stars 1 forks source link

model test #1

Closed shawrby closed 1 year ago

shawrby commented 1 year ago

First of all, thank you for your contribution!

How can I test your model?

andreasbinder commented 1 year ago

Hi! I will update the repository this weekend!

MaxistheSpy commented 1 year ago

Hi! I will update the repository this weekend!

does the model work, or is this project abandoned? Im looking into working with PointGNN and wanted a pytorch gemoetric version to work with, are there any instructions/readme?

andreasbinder commented 1 year ago

Hi @MaxistheSpy! Yes, this version is working I updated the model that I am using here.

@shawrby it expects the standard batch input from PyTorch-geometric:

import torch
from torch_geometric.data import Data, Batch

model = PointGNN_Normalization()

data_list = [Data(x=torch.tensor([[-1], [0], [1]], dtype=torch.float), 
            edge_index=torch.tensor([[0, 1, 1, 2], [1, 0, 2, 1]], dtype=torch.long))]

batch = Batch.from_data_list(data_list)

out = model(batch)

Does that work for you? Let me know if you encounter any problems! Note: if you just want to use the message passing layer, I released it to PyTorch-Geometric here I am sorry for the late reply!