---
license: apache-2.0
library_name: mlc-llm
tags:
  - askwise
  - prompt-engineering
  - qwen2
  - mlc
  - webgpu
base_model: Qwen/Qwen2.5-1.5B-Instruct
---

# AskWise-PromptEngineer-1.5B (MLC q4f16_1)

Fine-tuned **Qwen2.5-1.5B-Instruct** for AskWise Advanced rewrites:
messy user text → JSON `{"structured","advanced"}` prompt improvements (does **not** answer the task).

## Use with AskWise Chrome extension

These are **MLC WebLLM weights** (`q4f16_1`). The extension reuses its packaged
Qwen2.5-1.5B WebGPU wasm — only this weight folder is downloaded from Hugging Face.

## Converted with

- LoRA / QLoRA SFT on AskWise Instant seeds + synthetic prompt-engineering data
- `mlc_llm convert_weight … --quantization q4f16_1`