Milady Language Model

This model provides Milady-style responses to prompts, capturing the distinctive speech patterns and personality of Milady NFT characters.

Intended Use

The Milady Language Model is designed to:

• Generate text in the unique Milady style
• Create playful and creative responses to prompts
• Emulate Milady's distinctive personality

Usage

Run the model with vLLM:

docker run --gpus all \
  -p 8000:8000 \
  --ipc=host \
  vllm/vllm-openai:latest \
  --model theycallmeloki/milady \
  --tensor-parallel-size 4

Query the model:

curl -X POST http://localhost:8000/v1/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "theycallmeloki/milady",
    "prompt": "[INST] Transform the following prompt into a response written in Miladys characteristic style. Use playful language, creative spelling, and appropriate emojis to capture Miladys authentic voice and personality.\n\nYour question here [/INST]",
    "max_tokens": 200,
    "temperature": 0.9,
    "top_p": 0.95
  }'

This model was fine-tuned on a curated dataset of Milady-style text examples.

Additional Information

This model is based on the Mistral 7B Instruct v0.3 model, and is part of a greater collection of models that are part of the Milady Instrumentality Project.

This model provides 1-1 responses to a particular thought in the style of Milady. Subsequently the plan is to run a reasoning model and use this model to generate milady responses to thoughts generated from the reasoning model. Leading to synthetic training data for a girl math reasoning cute chain-of-thought milady model.