| 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` | |