R2G DATE Graph Dataset

R2G is a multi-view circuit graph benchmark suite from RTL to GDSII. This dataset repository provides the ready-to-train DATE graph artifacts for placement and routing tasks in both heterogeneous and homogeneous graph formats.

The graph artifacts are distributed as compressed archives containing PyTorch Geometric .pt files. See the R2G GitHub repository for data generation code, training scripts, and benchmark details:

• GitHub: https://github.com/ShenShan123/R2G
• arXiv: https://arxiv.org/abs/2604.08810
• GitHub Release mirror: https://github.com/ShenShan123/R2G/releases/tag/date-graph-dataset-v1.0

Files

data/
├── r2g-date-placement-heterogeneous-graphs.tar.gz
├── r2g-date-placement-homogeneous-graphs.tar.gz
├── r2g-date-routing-heterogeneous-graphs.tar.gz
└── r2g-date-routing-homogeneous-graphs.tar.gz

Each archive extracts into a stable DATE/ tree:

DATE/
├── placement/
│   ├── heterogeneous_graphs/
│   └── homogeneous_graphs/
└── routing/
    ├── heterogeneous_graphs/
    └── homogeneous_graphs/

See MANIFEST.tsv for the file-level mapping and SHA256SUMS for checksums.

Download

from huggingface_hub import snapshot_download

dataset_dir = snapshot_download(
    repo_id="yangaonjust/R2G",
    repo_type="dataset",
)
print(dataset_dir)

To download a single archive:

from huggingface_hub import hf_hub_download

path = hf_hub_download(
    repo_id="yangaonjust/R2G",
    repo_type="dataset",
    filename="data/r2g-date-placement-heterogeneous-graphs.tar.gz",
)
print(path)

After extracting an archive, load a graph artifact with PyTorch:

import torch

graph = torch.load("DATE/placement/heterogeneous_graphs/placement_view_b_heterogeneous_graph.pt")
print(graph)

Citation

If you use this dataset, please cite the R2G paper:

@misc{r2g2026,
  title = {R2G: A Multi-View Circuit Graph Benchmark Suite from RTL to GDSII},
  year = {2026},
  eprint = {2604.08810},
  archivePrefix = {arXiv}
}