File size: 2,103 Bytes
5683d95
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
---
license: apache-2.0
---

# CircuitNet 3.0 Dataset

This is the public dataset release for [CircuitNet 3.0: A Multi-Modal Dataset with Task-Oriented Augmentation for AI-Driven Circuit Design](https://openreview.net/forum?id=lEDb4gQ4dB).

The code and documentation companion lives in [sklp-eda-lab/iclr-circuitnet_3.0](https://github.com/sklp-eda-lab/iclr-circuitnet_3.0).

CircuitNet 3.0 is a multi-stage, multi-modal dataset for AI-driven circuit design. It is intended to support early-stage timing and power prediction, along with cross-stage representation learning over RTL, netlist, and layout-oriented artifacts.

## Current release

At the time of writing, the public archive is packaged as `circuitNetv3.zip`. The archive currently exposes:

- `dataset/RTL/` source RTL modules
- `dataset/Netlist/` synthesized netlist artifacts
- `dataset/Final/*/feature.json` gate-level feature metadata
- `dataset/Final/*/final_netlist.v` final synthesized netlists
- `dataset/Final/*/power_summary.txt` power summaries

Layout artifacts are large and are being uploaded incrementally. If a layout subset you need is not yet available, or if you would like to request a specific large layout package, please contact [wangmingjun000613@gmail.com](mailto:wangmingjun000613@gmail.com).

## Dataset summary

- 8,659 validated open-source RTL designs
- 15,863 total design instances after task-oriented augmentation
- Source-level train/validation/test partitioning, with augmented variants of test sources excluded from train/val to avoid leakage
- Timing labels: AT, WNS, and TNS
- Power labels: total power summaries

## What is intentionally excluded

- Commercial PDK files, standard-cell libraries, and other restricted technology collateral
- Proprietary EDA binaries or license-dependent artifacts
- Redundant examples that do not improve downstream use or reproducibility

## Update policy

This dataset card will be updated as new artifacts are uploaded. The Hugging Face hub is the source of truth for the latest available files.

## Citation

If you use this dataset, please cite the paper above.