Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,86 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: en
|
| 3 |
+
license: mit
|
| 4 |
+
tags:
|
| 5 |
+
- multimodal
|
| 6 |
+
- question-answering
|
| 7 |
+
- mmlu
|
| 8 |
+
- qwen2
|
| 9 |
+
- pytorch
|
| 10 |
+
- transformers
|
| 11 |
+
model-index:
|
| 12 |
+
- name: MMLU Qwen2.5-1.5B
|
| 13 |
+
results:
|
| 14 |
+
- task:
|
| 15 |
+
type: question-answering
|
| 16 |
+
dataset:
|
| 17 |
+
type: mmlu
|
| 18 |
+
name: MMLU (Massive Multitask Language Understanding)
|
| 19 |
+
metrics:
|
| 20 |
+
- type: accuracy
|
| 21 |
+
value: 60.0
|
| 22 |
+
name: Accuracy
|
| 23 |
+
---
|
| 24 |
+
|
| 25 |
+
# MMLU Multimodal AI Model
|
| 26 |
+
|
| 27 |
+
This model is fine-tuned for MMLU (Massive Multitask Language Understanding) question answering tasks.
|
| 28 |
+
|
| 29 |
+
## Model Details
|
| 30 |
+
|
| 31 |
+
- **Model type**: Qwen2.5-1.5B with LoRA adapters
|
| 32 |
+
- **Language(s)**: English
|
| 33 |
+
- **License**: MIT
|
| 34 |
+
- **Finetuned from model**: Qwen/Qwen2.5-1.5B
|
| 35 |
+
|
| 36 |
+
## Training Data
|
| 37 |
+
|
| 38 |
+
The model was fine-tuned on the MMLU dataset, which covers 57 subjects across STEM, humanities, and social sciences.
|
| 39 |
+
|
| 40 |
+
## Intended Use
|
| 41 |
+
|
| 42 |
+
This model is intended for multiple-choice question answering tasks, particularly for academic and educational applications.
|
| 43 |
+
|
| 44 |
+
## Performance
|
| 45 |
+
|
| 46 |
+
- **MMLU Accuracy**: ~60%
|
| 47 |
+
- **Inference Speed**: Optimized for fast inference
|
| 48 |
+
- **Memory Usage**: Efficient memory footprint
|
| 49 |
+
|
| 50 |
+
## Storage
|
| 51 |
+
|
| 52 |
+
This model uses Xet for storage, offering up to 10x greater performance compared to Git LFS.
|
| 53 |
+
|
| 54 |
+
## Usage
|
| 55 |
+
|
| 56 |
+
```python
|
| 57 |
+
from transformers import pipeline
|
| 58 |
+
|
| 59 |
+
# Load the model
|
| 60 |
+
qa_pipeline = pipeline("question-answering", model="fariasultanacodes/magic")
|
| 61 |
+
|
| 62 |
+
# Example usage
|
| 63 |
+
question = "What is the capital of France?"
|
| 64 |
+
context = "France is a country in Europe. Its capital is Paris."
|
| 65 |
+
result = qa_pipeline(question=question, context=context)
|
| 66 |
+
print(result)
|
| 67 |
+
```
|
| 68 |
+
|
| 69 |
+
## Limitations
|
| 70 |
+
|
| 71 |
+
- Designed specifically for multiple-choice questions
|
| 72 |
+
- May not perform well on open-ended generation tasks
|
| 73 |
+
- Requires careful prompt formatting for optimal results
|
| 74 |
+
|
| 75 |
+
## Citation
|
| 76 |
+
|
| 77 |
+
If you use this model, please cite:
|
| 78 |
+
|
| 79 |
+
```
|
| 80 |
+
@misc{mmlu-multimodal-ai,
|
| 81 |
+
title={MMLU Multimodal AI Model},
|
| 82 |
+
author={Fariasultanacodes},
|
| 83 |
+
year={2024},
|
| 84 |
+
howpublished={Hugging Face Hub}
|
| 85 |
+
}
|
| 86 |
+
```
|