Uploaded model

Developed by: kanoza
License: apache-2.0
Finetuned from model : unsloth/mistral-nemo-base-2407-bnb-4bit

This mistral model was trained 2x faster with Unsloth and Huggingface's TRL library.

Mistral Nemo MCQ Question Generator

Overview

A fine-tuned Mistral Nemo model specializing in generating multiple-choice questions (MCQs) across various domains.

Model Details

Base Model: Mistral Nemo Base 2407
Fine-Tuning: LoRA with 4-bit quantization
Training Dataset: SciQ
Primary Task: Automated MCQ Generation

Key Features

Scientific domain question generation
Supports multiple context types
High-quality, contextually relevant options
Configurable question complexity

Installation

pip install transformers unsloth
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("path/to/model")
tokenizer = AutoTokenizer.from_pretrained("path/to/model")

Usage Example

def generate_mcq(context, instruction):
    prompt = f"""
    Instruction: {instruction}
    Context: {context}
    """
    inputs = tokenizer(prompt, return_tensors="pt")
    outputs = model.generate(**inputs, max_new_tokens=128)
    return tokenizer.decode(outputs[0])

# Example application
context = "Photosynthesis converts sunlight into plant energy."
mcq = generate_mcq(context, "Create a multiple-choice question")
print(mcq)

Performance Metrics

BERTScore F1: [Placeholder]
ROUGE-1 F1: [Placeholder]
Generation Accuracy: [Placeholder]

Limitations

Primarily trained on scientific content
Requires careful prompt engineering
Potential bias in question generation

Ethical Considerations

Intended for educational research
Users should verify generated content

License

Apache 2.0

Contributing

Contributions welcome! Please open issues/PRs on GitHub.