---
library_name: transformers
tags:
  - mistral
  - instruct
  - causal-lm
  - sft
  - supervised-fine-tuning
  - hlx
  - mistral-7b-hlx2
  - personalized-model
  - conversational-ai
license: apache-2.0
language:
  - en
base_model: mistralai/Mistral-7B-Instruct-v0.2
datasets:
  - hailsbop/Haleyv2
pipeline_tag: text-generation
---

# Model Card for `mistral-7b-hlx2`

## Model Summary

`mistral-7b-hlx2` is a supervised fine-tuned (SFT) version of **Mistral-7B-Instruct-v0.2**, trained on **cleaned, de-identified WhatsApp conversation logs** (dataset: `hailsbop/Haleyv2`).  
The model is optimized for interpersonal dialogue, conversational tone, and expressive communication.

## Table of Contents
- Model Details
- Uses
- Bias, Risks, and Limitations
- Training Details
- Evaluation
- Societal Impact Assessment
- Model Examination
- Environmental Impact
- Technical Specifications
- Citation
- Glossary
- More Information
- Model Card Authors
- Model Card Contact
- How to Get Started

## Model Details

### Model Description
`mistral-7b-hlx2` is a decoder-only transformer model fine-tuned via SFT using WhatsApp-derived conversational data.  

- Developed by: Haley  
- License: Apache-2.0  
- Fine-tuned From: mistralai/Mistral-7B-Instruct-v0.2  
- Dataset: hailsbop/Haleyv2  

## Uses

### Direct Use
Conversational agents, expressive dialogue, creative writing.

### Downstream Use
Persona-aligned chatbots, stylistic generation.

### Out-of-Scope
Safety-critical use, factual decision-making, impersonation.

## Bias, Risks, and Limitations
This model reflects interpersonal tone, subjectivity, and informal WhatsApp communication patterns. Not suited for factual accuracy.

## Training Details

### Training Data
Dataset: `hailsbop/Haleyv2` — de-identified WhatsApp logs.

### Preprocessing
PII removed, paired into instruction–response sets.

### Training Procedure
Full SFT, mixed precision, AdamW optimizer.

## Evaluation
Qualitative testing only — no benchmarks.

## Environmental Impact
Training CO₂ impact not recorded.

## Technical Specifications
- 7B parameter decoder LM  
- FlashAttention-like optimizations  
- Causal LM objective  

## Citation
```
@misc{mistral7bhlx2,
  title={mistral-7b-hlx2},
  author={Haley},
  year={2025},
  howpublished={https://huggingface.co/hailsbop/mistral-7b-hlx2}
}
```

## Model Card Authors
Haley