README.md

---
license: other
license_name: deepseek
license_link: >-
  https://raw.githubusercontent.com/deepseek-ai/DeepSeek-Coder/main/LICENSE-MODEL
datasets:
- nllg/datikz-v2
---

# Model Card for DeTi*k*Zify-DS<sub>1.3b</sub>
DeTi*k*Zify-DS<sub>1.3b</sub> is a language model that automatically converts
sketches and existing scientific figures into editable, semantics-preserving
Ti*k*Z graphics programs. It is based on [DeepSeek Coder
1.3b](https://huggingface.co/deepseek-ai/deepseek-coder-1.3b-base) and was
fine-tuned on
[DaTi*k*Z<sub>v2</sub>](https://huggingface.co/datasets/nllg/datikz-v2). Check
out the [DeTi*k*Zify](https://github.com/potamides/DeTikZify) project for more
information and tips on how to best run the model.

## Usage
```python
from operator import itemgetter

from detikzify.model import load
from detikzify.infer import DetikzifyPipeline
import torch

image = "https://w.wiki/A7Cc"
pipeline = DetikzifyPipeline(*load(
    base_model="nllg/detikzify-ds-1.3b",
    device_map="auto",
    torch_dtype=torch.bfloat16,
))

# generate a single TikZ program
fig = pipeline.sample(image=image)

# if it compiles, rasterize it and show it
if fig.is_rasterizable:
    fig.rasterize().show()

# run MCTS for 10 minutes and generate multiple TikZ programs
figs = set()
for score, fig in pipeline.simulate(image=image, timeout=600):
    figs.add((score, fig))

# save the best TikZ program
best = sorted(figs, key=itemgetter(0))[-1][1]
best.save("fig.tex")
```