--- language: - zh - en library_name: peft pipeline_tag: text-generation base_model: Qwen/Qwen2.5-7B-Instruct tags: - peft - lora - qwen2.5 - reasoning - supervised-finetuning - text-generation - nycu-iaii-dl2026 --- # NYCU-IAII-DL2026 LLM2 SFT with Reasoning This repository contains the **LoRA adapter** trained for the NYCU-IAII-DL2026 LLM #2 task: **Reasoning LLM SFT with Reasoning Information**. The adapter is fine-tuned from: ```text Qwen/Qwen2.5-7B-Instruct ``` This repository does not contain a full merged model. It contains a PEFT / LoRA adapter that should be loaded on top of the base model. ## Model Details - Base model: Qwen/Qwen2.5-7B-Instruct - Fine-tuning method: Supervised fine-tuning with reasoning information - Adapter method: LoRA / PEFT - Quantization during training: 4-bit quantization - Framework: PyTorch, Transformers, PEFT - Task type: Multiple-choice question answering with reasoning information - Expected output: one of A, B, C, or D ## Files This repository includes: ```text adapter_config.json adapter_model.safetensors tokenizer.json tokenizer_config.json chat_template.jinja README.md ``` ## Limitations This adapter is trained specifically for the course multiple-choice reasoning task. It may not generalize well to open-ended reasoning or general chat scenarios.
