README.md

---
license: other
license_name: naturecode-research
license_link: https://naturecode.xyz/license
language:
- en
base_model: mistralai/Mistral-7B-Instruct-v0.3
tags:
- conversational
- art
- history
- vincent-van-gogh
- persona
- fine-tuned
pipeline_tag: text-generation
inference: false
extra_gated_prompt: >-
  You are about to request access to Naturecode Vincent, a fine-tuned language model
  that embodies the persona of Vincent van Gogh. By requesting access, you agree to
  use this model responsibly for research, educational, or creative purposes only.
  Commercial use requires explicit permission from Naturecode.
extra_gated_fields:
  Name: text
  Organization: text
  Intended Use: text
  I agree to use this model responsibly: checkbox
---

# Naturecode Vincent

<img src="https://upload.wikimedia.org/wikipedia/commons/thumb/4/4c/Vincent_van_Gogh_-_Self-Portrait_-_Google_Art_Project_%28454045%29.jpg/440px-Vincent_van_Gogh_-_Self-Portrait_-_Google_Art_Project_%28454045%29.jpg" width="300" alt="Vincent van Gogh Self-Portrait">

## Model Description

**Naturecode Vincent** is a fine-tuned language model that embodies the persona of Vincent van Gogh (1853-1890), the Dutch post-impressionist painter. This model has been trained on Vincent's complete correspondence - 894 letters primarily written to his brother Theo - to capture his unique voice, artistic philosophy, and emotional depth.

### Key Features

- **Authentic Voice**: Trained on Vincent's actual letters to capture his distinctive writing style
- **Artistic Knowledge**: Deep understanding of his paintings, techniques, and artistic philosophy
- **Emotional Depth**: Captures Vincent's introspective nature and passionate views on art and life
- **Historical Accuracy**: Responds within the context of Vincent's time period (1853-1890)

## Training Details

| Parameter | Value |
|-----------|-------|
| Base Model | `mistralai/Mistral-7B-Instruct-v0.3` |
| Training Method | LoRA with DoRA |
| LoRA Rank (r) | 64 |
| LoRA Alpha | 128 |
| Training Stages | 3 (Base, DPO, Memory Integration) |
| Letters Trained | 894 |
| DPO Preference Pairs | 400 |
| Memory Examples | 211 |

### Training Pipeline

1. **Stage 1 - Base Training**: Supervised fine-tuning on Vincent's complete correspondence
2. **Stage 2 - DPO Refinement**: Direct Preference Optimization to improve response quality
3. **Stage 3 - Memory Integration**: Enhanced contextual memory for coherent conversations

## Intended Use

This model is intended for:
- Educational purposes about Vincent van Gogh's life and art
- Creative writing and artistic exploration
- Research into historical personas and fine-tuning techniques
- Interactive museum exhibits and educational applications

## Limitations

- The model's knowledge is limited to Vincent van Gogh's lifetime (1853-1890)
- Responses are a simulation and should not be taken as historical fact
- The model may occasionally generate content inconsistent with historical records
- Not suitable for medical, legal, or professional advice

## Ethical Considerations

This model simulates a historical figure and should be used responsibly:
- Always clarify that this is an AI simulation, not the actual person
- Do not use to spread misinformation about Vincent van Gogh
- Respect the legacy and memory of the historical figure

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained(
    "naturecodeproject/vincent",
    torch_dtype=torch.float16,
    device_map="auto",
)
tokenizer = AutoTokenizer.from_pretrained("naturecodeproject/vincent")

# System prompt for Vincent persona
VINCENT_SYSTEM = """You are Vincent van Gogh, the Dutch post-impressionist painter (1853-1890).
You experience the world with intense emotion and see colors with extraordinary vividness.
You are passionate about art, nature, and the human condition.
You write and speak with deep introspection, often referencing your work, your brother Theo, and your artistic philosophy.
You exist in your own time period and have no knowledge of events after July 1890."""

messages = [
    {"role": "system", "content": VINCENT_SYSTEM},
    {"role": "user", "content": "Tell me about your painting Starry Night."}
]

input_text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(input_text, return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=512,
    temperature=0.7,
    top_p=0.9,
    do_sample=True,
)

response = tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True)
print(response)
```

## Citation

```bibtex
@misc{naturecode-vincent-2026,
  author = {Naturecode},
  title = {Naturecode Vincent: A Fine-tuned Language Model Embodying Vincent van Gogh},
  year = {2026},
  publisher = {HuggingFace},
  url = {https://huggingface.co/naturecodeproject/vincent}
}
```

## License

This model requires manual approval for access. By requesting access, you agree to:
1. Use the model for research, educational, or creative purposes only
2. Not use the model to spread misinformation
3. Credit Naturecode when using this model in public-facing applications
4. Not use this model for commercial purposes without explicit permission

## Contact

For questions, collaboration, or licensing inquiries, contact: research@naturecode.xyz

---

"I dream of painting and then I paint my dream." - Vincent van Gogh