dmlc/dgl

GINDataset is broken: dgl._ffi.base.DGLError: Expect argument "u" to have data type int32 or int64, but got torch.float32

Open

#5,749 opened on May 30, 2023

View on GitHub
 (11 comments) (0 reactions) (0 assignees)Python (12,665 stars) (2,928 forks)batch import
bug:unconfirmedhelp wanted

Description

🐛 Bug

This example breaks.

To Reproduce

[yuri@yv /usr/ports/math/py-dgl/work-py39/dgl-1.1.0/examples/pytorch/gin]$ python3.9 train.py --dataset MUTAG
Training with DGL built-in GINConv module with a fixed epsilon = 0
Traceback (most recent call last):
  File "/usr/ports/math/py-dgl/work-py39/dgl-1.1.0/examples/pytorch/gin/train.py", line 147, in <module>
    dataset = GINDataset(
  File "/usr/local/lib/python3.9/site-packages/dgl/data/gindt.py", line 132, in __init__
    super(GINDataset, self).__init__(
  File "/usr/local/lib/python3.9/site-packages/dgl/data/dgl_dataset.py", line 333, in __init__
    super(DGLBuiltinDataset, self).__init__(
  File "/usr/local/lib/python3.9/site-packages/dgl/data/dgl_dataset.py", line 112, in __init__
    self._load()
  File "/usr/local/lib/python3.9/site-packages/dgl/data/dgl_dataset.py", line 203, in _load
    self.process()
  File "/usr/local/lib/python3.9/site-packages/dgl/data/gindt.py", line 240, in process
    g.add_edges(j, nrow[2:])
  File "/usr/local/lib/python3.9/site-packages/dgl/heterograph.py", line 541, in add_edges
    u = utils.prepare_tensor(self, u, "u")
  File "/usr/local/lib/python3.9/site-packages/dgl/utils/checks.py", line 56, in prepare_tensor
    raise DGLError(
dgl._ffi.base.DGLError: Expect argument "u" to have data type int32 or int64, but got torch.float32.

dgl.data.GINDataset is broken. The above error is reproducible with:

from dgl.data import GINDataset

dataset = GINDataset("MUTAG", self_loop=True, degree_as_nlabel=False)

Environment

  • DGL Version (e.g., 1.0):1.1.0
  • Backend Library & Version (e.g., PyTorch 0.4.1, MXNet/Gluon 1.3): PyTorch-2.0.0
  • OS (e.g., Linux): FreeBSD 13.2
  • How you installed DGL (conda, pip, source): package
  • Build command you used (if compiling from source): n/a
  • Python version: 3.9
  • CUDA/cuDNN version (if applicable): n/a
  • GPU models and configuration (e.g. V100): n/a
  • Any other relevant information: numpy-1.23.1

Contributor guide

Tech stack
pythonpytorch
Domain
datamachine learning
Issue type
bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
mostly clear
Prerequisites
Basic understanding of DGLFamiliarity with PyTorch tensors
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
50
Research direction
Investigate the `gindt.py` file, specifically the `process` method where edges are added. The error occurs because `j` (a node index) is a float tensor instead of integer. Check the source of `j` and ensure it is cast to int32 or int64 before calling `add edges`. Look at linked PRs or comments for any proposed fixes.