curio184/qwen3-4b-struct-exp77
curio184/qwen3-4b-struct-exp77 is a 4 billion parameter causal language model, fine-tuned from Qwen/Qwen3-4B-Instruct-2507 using QLoRA. This model is specifically optimized to enhance structured output accuracy across various formats including JSON, YAML, XML, TOML, and CSV. It provides full merged 16-bit weights, eliminating the need for adapter loading, and is ideal for applications requiring precise data formatting.
Loading preview...
Model Overview
This model, curio184/qwen3-4b-struct-exp77, is a 4 billion parameter language model derived from Qwen/Qwen3-4B-Instruct-2507. It has been fine-tuned using QLoRA (4-bit, Unsloth) to significantly improve its ability to generate accurate structured data.
Key Capabilities
- Enhanced Structured Output: Specifically trained to produce precise JSON, YAML, XML, TOML, and CSV formats.
- Full Merged Weights: Distributed with full 16-bit merged weights, simplifying deployment as no adapter loading is required.
- Efficient Fine-tuning: Utilizes QLoRA for efficient training on a specialized dataset.
Training Details
The model was fine-tuned on the u-10bei/structured_data_with_cot_dataset_512_v2 dataset over 3 epochs. Key training parameters included a maximum sequence length of 512, a learning rate of 1e-06, and LoRA configuration with r=128 and alpha=256.
Ideal Use Cases
This model is particularly well-suited for applications where reliable and accurate structured data generation is critical, such as:
- API response generation
- Configuration file creation
- Data serialization tasks
- Automated report generation in structured formats