106 lines
3.0 KiB
Markdown
106 lines
3.0 KiB
Markdown
|
---
|
||
|
|
language:
|
||
|
|
- en
|
||
|
|
license: cc-by-nc-4.0
|
||
|
|
library_name: transformers
|
||
|
|
base_model: mistralai/Ministral-8B-Instruct-2410
|
||
|
|
base_model_relation: finetune
|
||
|
|
tags:
|
||
|
|
- conversational
|
||
|
|
- assistant
|
||
|
|
- fine-tuned
|
||
|
|
- lora
|
||
|
|
- collaborative
|
||
|
|
- vanta-research
|
||
|
|
- conversational-ai
|
||
|
|
- chat
|
||
|
|
- warm
|
||
|
|
- friendly-ai
|
||
|
|
- persona
|
||
|
|
- personality
|
||
|
|
- alignment
|
||
|
|
model-index:
|
||
|
|
- name: atom-v1-8b-preview
|
||
|
|
results: []
|
||
|
|
---
|
||
|
|
|
||
|
|
<div align="center">
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
<h1>VANTA Research</h1>
|
||
|
|
|
||
|
|
<p><strong>Independent AI safety research lab specializing in cognitive fit, alignment, and human-AI collaboration</strong></p>
|
||
|
|
|
||
|
|
<p>
|
||
|
|
<a href="https://unmodeledtyler.com"><img src="https://img.shields.io/badge/Website-unmodeledtyler.com-yellow" alt="Website"/></a>
|
||
|
|
<a href="https://x.com/vanta_research"><img src="https://img.shields.io/badge/@vanta_research-1DA1F2?logo=x" alt="X"/></a>
|
||
|
|
<a href="https://github.com/vanta-research"><img src="https://img.shields.io/badge/GitHub-vanta--research-181717?logo=github" alt="GitHub"/></a>
|
||
|
|
</p>
|
||
|
|
</div>
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
# Atom v1 8B Preview
|
||
|
|
|
||
|
|
Atom v1 8B Preview is a fine-tuned conversational AI model designed for collaborative problem-solving and thoughtful dialogue. Built on Mistral's Ministral-8B-Instruct-2410 architecture using Low-Rank Adaptation (LoRA), this model emphasizes natural engagement, clarifying questions, and genuine curiosity.
|
||
|
|
|
||
|
|
## Quick Start
|
||
|
|
|
||
|
|
```python
|
||
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM
|
||
|
|
|
||
|
|
model = AutoModelForCausalLM.from_pretrained("vanta-research/atom-v1-8b-preview", device_map="auto")
|
||
|
|
tokenizer = AutoTokenizer.from_pretrained("vanta-research/atom-v1-8b-preview")
|
||
|
|
|
||
|
|
messages = [
|
||
|
|
{"role": "system", "content": "You are Atom, a collaborative thought partner."},
|
||
|
|
{"role": "user", "content": "How do neural networks learn?"}
|
||
|
|
]
|
||
|
|
|
||
|
|
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to(model.device)
|
||
|
|
outputs = model.generate(inputs, max_new_tokens=512, temperature=0.8)
|
||
|
|
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
||
|
|
```
|
||
|
|
|
||
|
|
## Model Details
|
||
|
|
|
||
|
|
- **Developed by:** VANTA Research
|
||
|
|
- **Model type:** Causal language model
|
||
|
|
- **Base model:** mistralai/Ministral-8B-Instruct-2410
|
||
|
|
- **Parameters:** 8B
|
||
|
|
- **License:** CC BY-NC 4.0
|
||
|
|
- **Training method:** LoRA fine-tuning
|
||
|
|
- **Format:** Transformers (FP16) + GGUF (Q4_0)
|
||
|
|
|
||
|
|
## Capabilities
|
||
|
|
|
||
|
|
Optimized for:
|
||
|
|
- Collaborative problem-solving
|
||
|
|
- Technical explanations with accessible analogies
|
||
|
|
- Code generation and debugging
|
||
|
|
- Exploratory conversations
|
||
|
|
- Educational dialogue
|
||
|
|
|
||
|
|
## Files
|
||
|
|
|
||
|
|
- `*.safetensors` - Merged model weights (FP16)
|
||
|
|
- `atom-ministral-8b-q4_0.gguf` - Quantized model for Ollama/llama.cpp
|
||
|
|
- `config.json` - Model configuration
|
||
|
|
- `tokenizer.json` - Tokenizer files
|
||
|
|
|
||
|
|
## License
|
||
|
|
|
||
|
|
CC BY-NC 4.0 - Non-commercial use only. Contact VANTA Research for commercial licensing.
|
||
|
|
|
||
|
|
## Citation
|
||
|
|
|
||
|
|
```bibtex
|
||
|
|
@software{atom_v1_8b_preview,
|
||
|
|
title = {Atom v1 8B Preview},
|
||
|
|
author = {VANTA Research},
|
||
|
|
year = {2025},
|
||
|
|
url = {https://huggingface.co/vanta-research/atom-v1-8b-preview}
|
||
|
|
}
|
||
|
|
```
|