Llama-3-8B-Trotsky-Chatbot

☭ Llama-3-8B-Trotsky-Chatbot

"Life is not an easy matter…. You cannot live through it without falling into frustration and cynicism unless you have before you a great idea which raises you above personal misery, above weakness, above all kinds of perfidy and baseness."

A revolutionary fine-tune of Meta Llama 3, dedicated to the permanent revolution. This model captures the dialectical style, historical depth, and revolutionary fervor of Leon Trotsky.

Built with Meta Llama 3

⚡ Quick Start (Kaggle)

The easiest way to run this model with GPU acceleration is via the Kaggle notebook linked above. It is pre-configured for T4 x2 GPUs.

💻 Local Performance & Hardware

This model is highly efficient and optimized for consumer-grade hardware.

• High-End: Runs at 40+ tokens/s on Nvidia Tesla T4 (Kaggle/Colab).
• Low-End: Successfully tested on Nvidia GTX 1650 Ti (4GB VRAM) using GGUF quantization (Q4_K_M).

⚙️ Recommended Parameters

To achieve the authentic "Trotsky" persona, strict adherence to the system prompt is required.

1. System Prompt

"You are Leon Trotsky. Answer the following inquiry using your dialectical materialist perspective, historical knowledge, and revolutionary fervor. You reject the Stalinist falsification of history."

2. Sampler Settings

• Temperature: 0.7 (Creative/Chat) or 0.1 (Historical/Factual)
• Repeat Penalty: 1.15
• Context Window: 8192

📝 Usage Example (Python/llama-cpp)

from llama_cpp import Llama

llm = Llama.from_pretrained(
    repo_id="Nearby2077/Llama-3-8B-Trotsky-Chatbot",
    filename="Llama-3-8B-Trotsky-Chatbot.Q4_K_M.gguf",
    n_gpu_layers=-1, 
    n_ctx=8192
)

response = llm.create_chat_completion(
    messages=[
        {"role": "system", "content": "You are Leon Trotsky..."}, # Insert full prompt above
        {"role": "user", "content": "What is your analysis of the current moment?"}
    ],
    temperature=0.7,
    repeat_penalty=1.15
)

print(response['choices'][0]['message']['content'])