|
--- |
|
library_name: Llama3-8b-FlyingManual-Tutor |
|
tags: |
|
- llama3 |
|
- flying-manual |
|
- ai-tutoring |
|
- llama-factory |
|
--- |
|
|
|
# Model Card for Llama3-8b-FlyingManual-Tutor |
|
|
|
This model is a fine-tuned version of the Llama3-8b model, specifically trained on the FlyingManual dataset to serve as an AI tutor for aviation-related subjects. It is designed to provide guidance and nudge users when they answer questions incorrectly. |
|
|
|
## Model Details |
|
|
|
### Model Description |
|
|
|
- **Developed by:** Canarie Teams |
|
- **Model type:** Large Language Model (LLM) for AI Tutoring |
|
- **Language(s) (NLP):** English (primary), potentially others depending on the FlyingManual dataset |
|
- **Finetuned from model:** Llama3-8b by Meta AI |
|
|
|
## How to Get Started with the Model |
|
|
|
```python |
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
|
|
model_name = "path/to/your/Llama3-8b-FlyingManual-Tutor" |
|
tokenizer = AutoTokenizer.from_pretrained(model_name) |
|
model = AutoModelForCausalLM.from_pretrained(model_name) |
|
|
|
# Example usage for tutoring |
|
def tutor_interaction(question, user_answer): |
|
prompt = f"Question: {question}\nUser Answer: {user_answer}\nTutor Response:" |
|
inputs = tokenizer(prompt, return_tensors="pt") |
|
outputs = model.generate(**inputs, max_length=200) |
|
response = tokenizer.decode(outputs[0], skip_special_tokens=True) |
|
return response.split("Tutor Response:")[-1].strip() |
|
|
|
# Example |
|
question = "What are the primary flight controls of an aircraft?" |
|
user_answer = "Steering wheel and gas pedal" |
|
tutor_feedback = tutor_interaction(question, user_answer) |
|
print(tutor_feedback) |
|
``` |
|
|
|
## Training Details |
|
|
|
### Training Data |
|
|
|
The model was fine-tuned on the FlyingManual dataset, augmented with: |
|
- Sample Q&A pairs related to aviation topics |
|
- Examples of constructive feedback and explanations |
|
- Scenarios demonstrating correct and incorrect responses to aviation-related questions |
|
|
|
### Training Procedure |
|
|
|
#### Preprocessing |
|
|
|
- Conversion of training data into a dialogue format suitable for tutoring interactions |
|
- Augmentation of data with tutoring-specific tokens or markers |
|
|
|
#### Training Hyperparameters |
|
|
|
----------------------------- |
|
|
|
## Evaluation |
|
|
|
### Testing Data, Factors & Metrics |
|
|
|
#### Testing Data |
|
|
|
A held-out portion of the FlyingManual dataset, supplemented with: |
|
- A set of typical student questions and answers |
|
- Scenarios designed to test the model's ability to provide constructive feedback |
|
|
|
#### Metrics |
|
|
|
- Human evaluation of tutoring quality (clarity, accuracy, helpfulness) |
|
- Task-specific metrics (e.g., ability to correctly identify and address user mistakes) |
|
|
|
### Results |
|
|
|
[Provide the evaluation results here] |
|
|
|
## Environmental Impact |
|
|
|
- **Hardware Type:** 8 x NVIDIA A100 40GB GPUs |
|
|
|
|
|
## Model Card Authors |
|
|
|
Canarie Teams |
|
|
|
## Model Card Contact |
|
|
|
[Your contact information or a link to where people can reach out with questions] |
