alexsobolev/IcaroLM

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:1.5BQuant:BF16Ctx Length:32kPublished:Jun 15, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

IcaroLM by alexsobolev is a fine-tuned and quantized version of Qwen2 1.5B, specifically optimized for on-device mobile applications. With approximately 1.5 billion parameters and a quantized size of 600MB, it is designed for efficient local deployment on smartphones and edge devices. This model excels at maintaining emotionally intelligent conversations and executing reliable function calls within a chat flow, making it ideal for mobile assistants and emotional support applications.

Loading preview...

IcaroLM: Mobile-Optimized LLM for Empathetic Chat and Function Calling

IcaroLM, developed by alexsobolev, is a specialized fine-tuned and quantized version of the Qwen2 1.5B model. It is engineered for efficient on-device deployment, making it highly suitable for mobile and edge applications. The model's quantization reduces its footprint to approximately 600MB, enabling local inference without cloud dependency.

Key Capabilities

  • Mobile-Ready Footprint: Optimized for consumer mobile hardware with a compact 600MB size.
  • Function Calling: Explicitly fine-tuned to understand and execute function calls, facilitating local task automation and tool use.
  • Empathetic Chat: Trained on datasets focused on emotional intelligence for more natural and supportive conversational interactions.

Good for

  • Mobile Assistants: Creating local chatbots capable of performing device-specific actions via function calls.
  • Emotional Support Apps: Developing companion applications that require nuanced and empathetic conversational tones.
  • Edge Automation: Implementing task-oriented agents that operate locally with minimal latency.