AliesTaha/fable-traces
AliesTaha/fable-traces is a compact, instruction-tuned language model built upon the Qwen3-4B-Instruct-2507 architecture, featuring approximately 4 billion parameters. This model is specifically optimized for generating short, conversational replies, making it suitable for interactive applications. It is designed to run efficiently on a single mid-range GPU, offering accessibility for various deployment scenarios.
Loading preview...
Overview
AliesTaha/fable-traces is a compact, instruction-tuned language model derived from the Qwen3-4B-Instruct-2507 base model. With approximately 4 billion parameters and utilizing bfloat16 precision, it is engineered for efficient performance.
Key Capabilities
- Conversational Replies: The model is specifically tuned to generate short and concise responses, making it ideal for chat-based interactions.
- Resource Efficient: Designed to run comfortably on a single mid-range GPU, offering a cost-effective solution for deployment.
- ChatML Format: It uses the ChatML prompt format, leveraging the tokenizer's chat template for structured input.
Good For
- Applications requiring brief, interactive conversational outputs.
- Developers seeking a lightweight model that can be deployed on less powerful hardware.
- Use cases where fast inference for short text generation is critical.
Usage
The model can be easily integrated using the transformers library or served efficiently with vLLM for high-throughput inference.