tscircuit/zero-obstacle-high-density-z01

---
license: mit
dataset_info:
features:
- name: cond_image
dtype: image
- name: output_image
dtype: image
- name: instruction
dtype: string
splits:
- name: train
num_bytes: 11086366586
num_examples: 800000
- name: test
num_bytes: 2770961345
num_examples: 200000
download_size: 13690273789
dataset_size: 13857327931
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
task_categories:
- image-to-image
tags:
- electronics
- pcb
- autorouting
pretty_name: Zero Obstacle High Density Autorouting Dataset (Z01)
size_categories:
- 1M<n<10M
---

# tscircuit zero-obstacle-high-density-z01 Autorouting Dataset

A 1M sample synthetic autorouting dataset, created in collaboration with [seveibar](https://x.com/seveibar) and [makeshifted](https://huggingface.co/makeshifted)

[Blog Post](https://blog.autorouting.com/p/were-releasing-z01-1-million-autorouting) &middot; [Github](https://github.com/tscircuit/dataset-zero-obstacle-high-density-z01)

This dataset solves a subproblem of autorouting where there are 2 layers, no obstacles (pads, keepouts, traces) or impedance requirements. It represents the fundamental
problem of autorouting, a multi-agent pathfinding problem. A vision model that can solve this problem can directly integrate into the
[tscircuit autorouting pipeline](https://github.com/tscircuit/tscircuit-autorouter)

All outputs were generating using a grid-based rip'n'replace autorouter [@tscircuit/high-density-a01](https://github.com/tscircuit/high-density-a01)

This dataset is part of the [tscircuit](https://tscircuit.com) and [autorouting.com](https://blog.autorouting.com) projects to
create a fast modern autorouter.

[Join us on Discord](https://tscircuit.com/join) &middot; [tscircuit](https://tscircuit.com)