84basi/lora-10-1
TEXT GENERATIONConcurrency Cost:1Model Size:4BQuant:BF16Ctx Length:32kPublished:Mar 1, 2026License:apache-2.0Architecture:Transformer Open Weights Warm
The 84basi/lora-10-1 model is a 4 billion parameter Qwen3-based language model, fine-tuned from unsloth/Qwen3-4B-Instruct-2507 using BF16 full fine-tuning and NEFTune. It is specifically optimized for generating accurate structured output in formats like JSON, YAML, XML, TOML, and CSV. This model excels at tasks requiring precise data formatting by removing Chain-of-Thought from its training data.
Loading preview...
Model Overview
84basi/lora-10-1 is a 4 billion parameter language model derived from unsloth/Qwen3-4B-Instruct-2507. It has undergone full fine-tuning using BF16 precision and NEFTune, with a focus on enhancing its ability to produce structured outputs.
Key Capabilities
- Structured Output Generation: Specifically trained to improve accuracy when generating data in formats such as JSON, YAML, XML, TOML, and CSV.
- Optimized for Direct Output: Chain-of-Thought (CoT) reasoning was intentionally removed from the training data to promote direct and precise structured responses.
- Training Details: Fine-tuned for 1 epoch with a learning rate of 2e-05 and a NEFTune noise alpha of 5.0, using a maximum sequence length of 2048 tokens.
Good For
- Applications requiring reliable and accurate JSON, YAML, XML, TOML, or CSV output.
- Use cases where a model needs to directly output structured data without intermediate reasoning steps.
- Integration into systems that parse machine-readable data directly from model responses.