Hardman/upsc-question-generator

UPSC Question Generator (Granite 350M, merged 4-bit)

Fine-tuned model for generating UPSC Civil Services Mains-style questions (GS1–GS4).

Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer model = AutoModelForCausalLM.from_pretrained("Hardman/upsc-question-generator") tokenizer = AutoTokenizer.from_pretrained("Hardman/upsc-question-generator") prompt = "<|user|> Generate 3 UPSC questions on Indian Economy <|assistant|>" inputs = tokenizer(prompt, return_tensors="pt") outputs = model.generate(**inputs, max_new_tokens=512, temperature=0.7, do_sample=True) print(tokenizer.decode(outputs, skip_special_tokens=True))

text

Inference API

import requests API_URL = "https://router.huggingface.co/hf-inference/models/{HF_USERNAME}/upsc-question-generator" headers = {"Authorization": "Bearer YOUR_HF_TOKEN"} resp = requests.post( API_URL, headers=headers, json={ "inputs": "<|user|> Generate UPSC questions on Science & Technology <|assistant|>", "parameters": {"max_new_tokens": 512, "temperature": 0.7, "top_p": 0.9, "do_sample": True} } ) print(resp.json())

text

Training Notes

• Base: IBM Granite 3.0 350M Instruct
• Method: LoRA → merged into 4-bit base
• Data: 1000+ UPSC Mains questions (GS1–GS4)
• Hardware: Colab T4
• Steps: ~200

License

Apache 2.0