torch-geometric
Graph Neural Networks (PyG). Node/graph classification, link prediction, GCN, GAT, GraphSAGE, heterogeneous graphs, molecular property prediction, for geometric deep learning.
Security Assessment
Detected risks:
About torch-geometric
The torch-geometric skill provides a comprehensive framework for developing and training Graph Neural Networks (GNNs) using PyTorch. This library is specifically designed to handle graph-based data structures, allowing users to work effectively with irregular data. By leveraging mini-batch processing and enabling multi-GPU training, torch-geometric addresses the challenges of scaling GNNs for large datasets and complex models, making it a powerful tool for researchers and practitioners in the field of geometric deep learning.
FAQ
How do I install torch-geometric?
You can install torch-geometric using the command: `pip install torch_geometric`. For additional dependencies, use: `pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-${TORCH}+${CUDA}.html`.
Can I use torch-geometric for molecular property prediction?
Yes, torch-geometric is well-suited for molecular property prediction and can be applied in drug discovery and chemical property prediction tasks.
What types of graph structures are supported?
torch-geometric supports various graph structures including heterogeneous graphs with multi-type nodes and edges, as well as large-scale graphs.
Are there any specific hardware requirements for using this skill?
While torch-geometric can run on CPUs, it is recommended to use a multi-GPU setup for better performance, especially for training large models.
What are the limitations of torch-geometric?
Some limitations include the need for custom adaptations for very specific graph types and the dependency on PyTorch, which means it may not be compatible with other deep learning frameworks.
Install torch-geometric
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
davila7/claude-code-templates