ermiaazarkhalili/Qwen3.8-9B-Function-Calling-xLAM-Unsloth

Model Overview

This model, ermiaazarkhalili/Qwen3.8-9B-Function-Calling-xLAM-Unsloth, is a LoRA fine-tune of the empero-ai/Qwen3.8-9B base model, featuring 9.7 billion parameters. It was specifically trained using the Unsloth and TRL frameworks on the Salesforce/xlam-function-calling-60k dataset, focusing on supervised fine-tuning for function calling capabilities. The training involved a single epoch with a maximum sequence length of 2048 and utilized 4-bit QLoRA precision.

Key Capabilities

• Function Calling: Specialized in generating structured function calls based on input instructions, making it suitable for integrating with external tools and APIs.
• Qwen3.8 Architecture: Built upon the Qwen3.8-9B base model, inheriting its general language understanding and generation abilities.
• Efficient Fine-tuning: Leverages LoRA (r=64, alpha=64) and Unsloth for efficient adaptation, allowing for effective specialization without full model retraining.

Limitations and Considerations

• No Benchmark Evaluation: The model has not undergone downstream benchmark evaluation; reported numbers are solely training-loss observations.
• Inherited Biases: It carries the biases, knowledge cutoff, and potential failure modes of its base model.
• Specialized Behavior: Fine-tuned on a single instruction-following dataset, its performance outside this specific distribution is untested.
• Merged Adapters: The LoRA adapters are merged into the base weights, meaning the fine-tune cannot be detached from the base model.