SecretLab/trump-mistral-adapter

<div align="center">

Trump Mistral Adapter

<img src="https://img.shields.io/badge/MODEL-Mistral--7B-blue?style=for-the-badge" alt="Model: Mistral-7B"/> <img src="https://img.shields.io/badge/ADAPTER-LoRA-red?style=for-the-badge" alt="Adapter: LoRA"/> <img src="https://img.shields.io/badge/STYLE-Trump-yellow?style=for-the-badge" alt="Style: Trump"/>

</div>

"This adapter, believe me folks, it's tremendous. It's the best adapter, everyone says so. We're going to do things with this model that nobody's ever seen before."

A fine-tuned language model that captures Donald Trump's distinctive speaking style, discourse patterns, and policy positions. This LoRA adapter transforms Mistral-7B-Instruct-v0.2 to emulate the unique rhetorical flourishes and speech cadence of the former U.S. President.

<div align="center"> <img src="https://img.shields.io/badge/🗣️ Speech Patterns-✓-success" alt="Speech Patterns"/> <img src="https://img.shields.io/badge/🏛️ Policy Positions-✓-success" alt="Policy Positions"/> <img src="https://img.shields.io/badge/🔄 Repetition Style-✓-success" alt="Repetition Style"/> <img src="https://img.shields.io/badge/👐 Hand Gestures-✗-lightgrey" alt="Hand Gestures"/> </div>

🔍 Overview

🚀 Getting Started

💻 Python Implementation

from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
from peft import PeftModel
import torch

# Configuration
base_model_id = "mistralai/Mistral-7B-Instruct-v0.2"
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_compute_dtype=torch.float16,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_use_double_quant=True,
)

# Load model and tokenizer
model = AutoModelForCausalLM.from_pretrained(
    base_model_id,
    quantization_config=bnb_config,
    device_map="auto",
    torch_dtype=torch.float16
)
tokenizer = AutoTokenizer.from_pretrained(base_model_id)

# Apply adapter
model = PeftModel.from_pretrained(model, "nnat03/trump-mistral-adapter")

# Generate response
prompt = "What's your plan for border security?"
input_text = f"<s>[INST] {prompt} [/INST]"
inputs = tokenizer(input_text, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_length=512, temperature=0.7, do_sample=True)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response.split("[/INST]")[-1].strip())

🔮 Ollama Integration

For simplified local deployment:

# Pull the model
ollama pull nnat03/trump-mistral

# Run the model
ollama run nnat03/trump-mistral

Access this model via the Ollama library.

📊 Example Output

<table> <tr> <th width="20%">Topic</th> <th>Response</th> </tr> <tr> <td>Border Security</td> <td><i>"First of all, we need the wall. The wall is very important. It's not just a wall, it's steel and concrete and things that are very, very strong. We have 450 miles completed. It's an incredible job."</i></td> </tr> <tr> <td>Joe Biden</td> <td><i>"Joe Biden, I call him 1% Joe. His numbers are way down. He's a corrupt politician. He's been there for 47 years. Where has he been? What's he done? There's nothing."</i></td> </tr> </table>

⚙️ Technical Details

📚 Training Data

This model was trained on authentic speech patterns from:

• —Trump interviews dataset (pookie3000/trump-interviews)
• —Trump speeches dataset (bananabot/TrumpSpeeches)

🔧 Model Configuration

LoRA rank: 16 (tremendous rank, the best rank)
Alpha: 64
Dropout: 0.05
Target modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj

🧠 Training Parameters

Batch size: 4
Gradient accumulation: 4
Learning rate: 2e-4
Epochs: 3
LR scheduler: cosine
Optimizer: paged_adamw_8bit
Precision: BF16

🎯 Applications

<div align="center"> <table width="80%"> <tr> <td align="center" width="33%"><b>🎓 Education</b><br><small>Political discourse analysis</small></td> <td align="center" width="33%"><b>🔬 Research</b><br><small>Rhetoric pattern studies</small></td> <td align="center" width="33%"><b>🎭 Creative</b><br><small>Interactive simulations</small></td> </tr> </table> </div>

⚠️ Notes and Limitations

This model mimics a speaking style but does not guarantee factual accuracy or represent actual views. It may reproduce biases present in the training data and is primarily intended for research and educational purposes.

📄 Citation

@misc{nnat03-trump-mistral-adapter,
  author = {nnat03},
  title = {Trump Mistral Adapter},
  year = {2023},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/nnat03/trump-mistral-adapter}}
}

<div align="center"> <p><b>Framework version:</b> PEFT 0.15.0</p> <p>Created for NLP research and education</p> <p><small>"We're gonna have the best models, believe me."</small></p> </div>