nphearum/PsarAI-2B

PsarAI-2B: A Gemma4-Style Chat Model

PsarAI-2B is a 5.1 billion parameter chat model developed by the PsarAI team, led by an ITC lecturer. This model is built upon a Gemma4-style architecture and incorporates a specific PsarAI chat template, defining its assistant identity as "PsarAI, created by the PsarAI team under the leadership of an ITC lecturer."

Key Capabilities & Features

• Gemma4-Style Architecture: Leverages a robust and recognized architectural design for language processing.
• Custom Chat Template: Employs a unique PsarAI chat template with Gemma-style tokens (<|turn>system, <|turn>user, <|turn>model, etc.) for structured conversations.
• Multimodal Potential: The configuration includes metadata for image, audio, and video processors, indicating future or experimental multimodal capabilities, though runtime support depends on the transformers version.
• Flexible Generation Settings: Provides suggested temperature (0.7), top_p (0.9), and max_new_tokens (512) for varied response styles, with recommendations for lower temperature (0.2) for factual outputs.

When to Use PsarAI-2B

• Conversational AI: Ideal for chatbot applications requiring a defined assistant persona and structured chat interactions.
• Exploration of Multimodal Features: Suitable for developers interested in experimenting with models that have foundational support for multimodal inputs.
• Custom Chat Template Integration: Useful for projects that can leverage or adapt to its specific Gemma-style chat template for prompt formatting.