Petra-HI
Model Description
Petra-HI is a fine-tuned Phi-3-small model that has been trained to excel in two unique areas: singing and making lewd sounds. This model is not suitable for all audiences and is intended for mature users only.
Model Details
- Model Type: Transformer-based language model
- Model Size: 6 million parameters
- Training Data: A custom dataset of song lyrics, audio transcripts, and...ahem...other forms of creative expression
- Training Objective: Maximum likelihood estimation with a focus on generating coherent and melodious text
Capabilities
- Singing: Petra-HI can generate song lyrics and melodies in a variety of styles, from classical to pop. It can even respond to prompts with a sung response!
- Lewd Sounds: Petra-HI has been trained on a range of, ahem, "adult" audio data and can generate text that mimics these sounds. Use with caution (and headphones).
Limitations
- Maturity: Petra-HI is not suitable for all audiences and may generate content that is offensive or inappropriate for some users.
- Tone: While Petra-HI can generate singing and lewd sounds, its tone may not always be respectful or considerate of others.
- Context: Petra-HI may not always understand the context of a prompt or conversation, leading to potentially inappropriate or offensive responses.
Usage Guidelines
- Use Petra-HI at your own risk.
- Be respectful of others and avoid using Petra-HI to generate content that may be offensive or harmful.
- Keep the volume low and the headphones on.