ermiaazarkhalili/Qwen3-4B-Function-Calling-xLAM-Unsloth
The ermiaazarkhalili/Qwen3-4B-Function-Calling-xLAM-Unsloth model is a 4 billion parameter Qwen3-based causal language model, fine-tuned for function calling capabilities. Developed by ermiaazarkhalili, it leverages Unsloth for efficient training, resulting in 2x faster fine-tuning and 60% less VRAM usage. This model excels at interpreting natural language queries and generating structured function calls, making it suitable for integrating LLMs with external tools and APIs.
Loading preview...
Overview
This model, developed by ermiaazarkhalili, is a 4 billion parameter Qwen3-based language model specifically fine-tuned for function calling. It utilizes the Unsloth framework, which enabled 2x faster training and significantly reduced VRAM consumption (60% less) during its development. The base model is Qwen3-4B (Unsloth 4-bit).
Key Capabilities & Training
- Function Calling Specialization: Optimized to understand user queries and generate appropriate function calls, trained on the Salesforce/xlam-function-calling-60k dataset, which contains 60,000 examples.
- Efficient Fine-tuning: Benefits from Unsloth's optimizations, including SFT with QLoRA (4-bit) and an optimized gradient checkpointing, achieving a final training loss of 0.2309 in under 2 hours on an NVIDIA H100 GPU.
- Compact & Performant: Despite its 4B parameters, its function-calling specialization makes it efficient for specific tasks. The model was fine-tuned with a 2,048 token context window.
- Versatile Deployment: Available in various formats, including standard Transformers for GPU inference, Unsloth for fastest inference, and GGUF versions (Q4_K_M, Q5_K_M, Q8_0) for CPU and edge devices via Ollama or llama.cpp.
Ideal Use Cases
- Tool Integration: Excellent for applications requiring an LLM to interact with external APIs, databases, or services by generating structured function calls.
- Automated Workflows: Suitable for building agents that can perform actions based on user commands.
- Resource-Constrained Environments: The Unsloth optimization and GGUF availability make it viable for deployment on hardware with limited resources, including CPU-only setups.