Update README.md
Browse files
README.md
CHANGED
@@ -85,4 +85,67 @@ library_name: transformers
|
|
85 |
|
86 |
This mistral model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
|
87 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
88 |
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
|
|
|
85 |
|
86 |
This mistral model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
|
87 |
|
88 |
+
# Mistral Nemo MCQ Question Generator
|
89 |
+
|
90 |
+
## Overview
|
91 |
+
A fine-tuned Mistral Nemo model specializing in generating multiple-choice questions (MCQs) across various domains.
|
92 |
+
|
93 |
+
## Model Details
|
94 |
+
- **Base Model**: Mistral Nemo Base 2407
|
95 |
+
- **Fine-Tuning**: LoRA with 4-bit quantization
|
96 |
+
- **Training Dataset**: SciQ
|
97 |
+
- **Primary Task**: Automated MCQ Generation
|
98 |
+
|
99 |
+
## Key Features
|
100 |
+
- Scientific domain question generation
|
101 |
+
- Supports multiple context types
|
102 |
+
- High-quality, contextually relevant options
|
103 |
+
- Configurable question complexity
|
104 |
+
|
105 |
+
## Installation
|
106 |
+
```python
|
107 |
+
pip install transformers unsloth
|
108 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
109 |
+
|
110 |
+
model = AutoModelForCausalLM.from_pretrained("path/to/model")
|
111 |
+
tokenizer = AutoTokenizer.from_pretrained("path/to/model")
|
112 |
+
```
|
113 |
+
|
114 |
+
## Usage Example
|
115 |
+
```python
|
116 |
+
def generate_mcq(context, instruction):
|
117 |
+
prompt = f"""
|
118 |
+
Instruction: {instruction}
|
119 |
+
Context: {context}
|
120 |
+
"""
|
121 |
+
inputs = tokenizer(prompt, return_tensors="pt")
|
122 |
+
outputs = model.generate(**inputs, max_new_tokens=128)
|
123 |
+
return tokenizer.decode(outputs[0])
|
124 |
+
|
125 |
+
# Example application
|
126 |
+
context = "Photosynthesis converts sunlight into plant energy."
|
127 |
+
mcq = generate_mcq(context, "Create a multiple-choice question")
|
128 |
+
print(mcq)
|
129 |
+
```
|
130 |
+
|
131 |
+
## Performance Metrics
|
132 |
+
- BERTScore F1: [Placeholder]
|
133 |
+
- ROUGE-1 F1: [Placeholder]
|
134 |
+
- Generation Accuracy: [Placeholder]
|
135 |
+
|
136 |
+
## Limitations
|
137 |
+
- Primarily trained on scientific content
|
138 |
+
- Requires careful prompt engineering
|
139 |
+
- Potential bias in question generation
|
140 |
+
|
141 |
+
## Ethical Considerations
|
142 |
+
- Intended for educational research
|
143 |
+
- Users should verify generated content
|
144 |
+
|
145 |
+
## License
|
146 |
+
Apache 2.0
|
147 |
+
|
148 |
+
## Contributing
|
149 |
+
Contributions welcome! Please open issues/PRs on GitHub.
|
150 |
+
|
151 |
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
|