🧠 SocraticLM-Qwen-LoRA

LoRA-Fine-Tuned Qwen2.5-7B-Instruct for Pedagogy-Optimized Teaching Conversations

This repository contains a LoRA adapter fine-tuned on 160k curated pedagogical dialogs
designed for step-by-step Socratic teaching, self-explanation, and guided reasoning.

The base model remains unchanged:
👉 Qwen/Qwen2.5-7B-Instruct
and this repo stores only the LoRA weights (adapter_model.safetensors).

---

🚀 How to Use

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base = "Qwen/Qwen2.5-7B-Instruct"
adapter = "Aditya-m04/SocraticLM-Qwen-Lora"

tokenizer = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(base, device_map="auto")
model = PeftModel.from_pretrained(model, adapter)

inp = "Explain the Pythagorean theorem to a 10-year-old."
out = model.generate(**tokenizer(inp, return_tensors="pt").to(model.device), max_new_tokens=200)
print(tokenizer.decode(out[0], skip_special_tokens=True))