ermiaazarkhalili/Qwen2.5-1.5B-Instruct_Function_Calling_xLAM

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:1.5BQuant:BF16Ctx Length:32kPublished:Aug 1, 2025License:apache-2.0Architecture:Transformer Open Weights Warm

ermiaazarkhalili/Qwen2.5-1.5B-Instruct_Function_Calling_xLAM is a 1.5 billion parameter language model developed by ermiaazarkhalili, fine-tuned from Qwen/Qwen2.5-1.5B-Instruct. This model is specifically optimized for function calling tasks, leveraging Supervised Fine-Tuning (SFT) with LoRA adapters on the Salesforce/xlam-function-calling-60k dataset. It is designed for efficient inference and research into fine-tuning methods for instruction-following models.

Loading preview...

Model Overview

This model, developed by ermiaazarkhalili, is a 1.5 billion parameter language model derived from Qwen/Qwen2.5-1.5B-Instruct. It has been specifically fine-tuned using Supervised Fine-Tuning (SFT) with LoRA adapters on the Salesforce/xlam-function-calling-60k dataset, making it specialized for function calling tasks.

Key Capabilities

  • Function Calling Optimization: Fine-tuned on a dedicated function-calling dataset to improve instruction following for tool use.
  • Efficient Training: Utilizes LoRA (Low-Rank Adaptation) with 4-bit quantization for efficient fine-tuning.
  • Base Model: Built upon the Qwen2.5-1.5B-Instruct architecture, providing a strong foundation for instruction adherence.
  • Inference Flexibility: Available in various formats, including GGUF quantizations for CPU or mixed CPU/GPU inference.

Intended Use Cases

  • Research: Ideal for exploring language model fine-tuning techniques and function calling capabilities.
  • Educational Purposes: Suitable for learning and experimentation with LLMs and SFT.
  • Prototyping: Useful for developing and testing conversational AI agents that require function calling.
  • Personal Projects: Applicable for various personal applications where a compact, function-calling capable model is beneficial.

Limitations

  • Primarily trained on English data.
  • Fine-tuned with a 2,048 token context length.
  • May exhibit hallucinations and is not extensively safety-tuned.