NullvoidXO/elegoo-resin-printing-v1

---
license: apache-2.0
dataset_info:
license: apache-2.0
---

# Elegoo Resin Printing Dataset v1

A specialized dataset for fine-tuning language models on 3D resin printing with Elegoo printers.

## Description

This dataset contains training examples for building AI assistants that can:
- Recommend optimal slicer settings (exposure, lift speed, layer height) for specific printer/resin combinations
- Diagnose print failures and provide actionable fixes
- Generate ChiTuBox/Lychee slicer configuration files

## Dataset Composition

| Source | Count | Description |
|--------|-------|-------------|
| Synthetic (settings) | 2,447 | Slicer settings recommendations grounded in scraped docs |
| Synthetic (diagnosis) | 2,435 | Failure diagnosis Q&A |
| Synthetic (config) | 2,393 | Slicer config file generation (ChiTuBox/Lychee) |
| Synthetic (Mercury) | 143 | Wash/cure station diagnosis (Mercury Plus V2 focus) |
| Real Q&A | 109 | StackExchange resin printing questions |
| Doc chunks | 344 | Converted documentation excerpts |
| **Total** | **7,869** | |

## Priority Coverage

A special focus was applied to two of the most commonly used products:
- **Mars 4 Ultra**: 3x weight in synthetic generation (~18% of printer mentions)
- **Mercury Plus V2**: 150 diagnosis examples (wash/cure failures)

## Format

Conversational format compatible with SFT training:

```
### Human:
{instruction/input}

### Assistant:
{output}
```

## Intended Use

Fine-tuning language models (e.g., Qwen3.5-9B, Llama 3.1 8B) for:
- Resin printing assistance
- Slicer settings optimization
- Print failure diagnosis

## Data Sources

### Scraped (used for grounding)
- Elegoo Wiki (wiki.elegoo.com)
- ChiTuBox Documentation (docs.chitubox.com)
- Lychee/Mango3D Documentation (docs.mango3d.io)
- 3DPrinting StackExchange (via HuggingFace dataset)

### Synthetic Generation
Generated using Ollama Cloud API (qwen3-coder-next:cloud) with doc-grounded prompting.

## License

- Synthetic data: Apache 2.0
- Scraped documentation: Original licenses retained (Elegoo, ChiTuBox, Mango3D)
- StackExchange: CC BY-SA 4.0

## Training Script

Compatible with standard SFT pipelines:

```python
from datasets import load_dataset
from transformers import AutoTokenizer

dataset = load_dataset("your-username/elegoo-resin-printing-v1")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3.5-9B")

def tokenize(example):
return tokenizer(example["text"], truncation=True, max_length=2048)

tokenized = dataset.map(tokenize)
```

## Version History

- **v1** (2026-04-17): Initial release, 7,869 examples
- Text-based V1 scope (settings, diagnosis, config)
- V2 features deferred (image-to-3D, YouTube transcripts)

## Citation

```bibtex
@dataset{elegoo-resin-printing-v1,
author = {Your Name},
title = {Elegoo Resin Printing Dataset v1},
year = {2026},
url = {https://huggingface.co/datasets/your-username/elegoo-resin-printing-v1}
}
```