erickgm1998/mermaid-qwen-0.5b-onnx

Mermaid Qwen 0.5B ONNX

Fine-tuned Qwen2.5 0.5B model for generating Mermaid diagrams.

Base Model

Qwen/Qwen2.5-0.5B-Instruct

Runtime

This repository is structured for @huggingface/transformers / Transformers.js.

Available ONNX Files

• onnx/model.onnx + onnx/model.onnx_data: FP32 export.
• onnx/model_fp16.onnx: FP16 export.
• onnx/model_quantized.onnx: Q8 alias used by Transformers.js.
• onnx/model_q8.onnx: Q8 export.
• onnx/model_q4.onnx: Q4 export.

Recommended First Test

Start with dtype: "fp32" for quality validation. The quantized variants should be tested carefully for Mermaid syntax quality.