|
--- |
|
library_name: transformers |
|
tags: [] |
|
--- |
|
|
|
# Model Card for LLaMAdelic |
|
|
|
<!-- Summary of model purpose and objective --> |
|
The LLaMAdelic model was fine-tuned to generate conversations with personality-specific adaptations based on the Big 5 personality traits: Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism. The model was trained on conversational data, specifically designed to enhance its capacity for personality-driven dialogue. Detailed information about the dataset will be updated soon. |
|
|
|
## Model Details |
|
Will be updated soon |
|
|
|
### Model Description |
|
|
|
This is the model card for the fine-tuned LLaMAdelic model on the 🤗 Hub. |
|
|
|
- **Developed by:** Will be updated soon |
|
- **Funded by [optional]:** Will be updated soon |
|
- **Shared by [optional]:** Will be updated soon |
|
- **Model type:** Fine-tuned LLaMA 3 8B Instruct |
|
- **Language(s) (NLP):** English |
|
- **License:** Will be updated soon |
|
- **Finetuned from model:** [LLaMA 3 8B Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct) |
|
|
|
### Model Sources |
|
|
|
- **Repository:** Will be updated soon |
|
- **Paper [optional]:** Will be updated soon |
|
- **Demo [optional]:** Will be updated soon |
|
|
|
## Uses |
|
|
|
### Direct Use |
|
Will be updated soon |
|
|
|
### Downstream Use |
|
Will be updated soon |
|
|
|
### Out-of-Scope Use |
|
Will be updated soon |
|
|
|
## Bias, Risks, and Limitations |
|
Will be updated soon |
|
|
|
### Recommendations |
|
Users (both direct and downstream) should be aware of potential biases and limitations. Further recommendations will be updated soon. |
|
|
|
## How to Get Started with the Model |
|
Will be updated soon |
|
|
|
## Training Details |
|
|
|
### Training Data |
|
Will be updated soon |
|
|
|
### Training Procedure |
|
|
|
#### Preprocessing |
|
Will be updated soon |
|
|
|
#### Training Hyperparameters |
|
Will be updated soon |
|
|
|
#### Speeds, Sizes, Times |
|
Will be updated soon |
|
|
|
## Evaluation |
|
|
|
### Testing Data, Factors & Metrics |
|
|
|
#### Testing Data |
|
Will be updated soon |
|
|
|
#### Factors |
|
Will be updated soon |
|
|
|
#### Metrics |
|
Will be updated soon |
|
|
|
### Results |
|
|
|
The results from the [EleutherAI/lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness) for OCEAN personality benchmarks show notable improvements in personality-specific scores for the LLaMAdelic model over the zeroshot pre-trained LLaMA 8B instruct model. Detailed evaluation results are below: |
|
|
|
| Model | Description | Openness | Conscientiousness | Extraversion | Agreeableness | Neuroticism | Avg | |
|
|------------------|------------------------------------|----------|-------------------|--------------|---------------|-------------|------| |
|
| LLaMA 8B ins | Zeroshot | 0.8760 | 0.7620 | 0.7170 | 0.9500 | 0.5220 |0.7654| |
|
| LLaMAdelic | Finetuned on Conversational Data | 0.9150 | 0.7840 | 0.6680 | 0.9440 | 0.7040 |0.8030| |
|
|
|
## Model Examination |
|
Will be updated soon |
|
|
|
## Environmental Impact |
|
Will be updated soon |
|
|
|
## Technical Specifications |
|
|
|
### Model Architecture and Objective |
|
Will be updated soon |
|
|
|
### Compute Infrastructure |
|
Will be updated soon |
|
|
|
#### Hardware |
|
Will be updated soon |
|
|
|
#### Software |
|
Will be updated soon |
|
|
|
## Citation |
|
Will be updated soon |
|
|
|
## Glossary |
|
Will be updated soon |
|
|
|
## More Information |
|
Will be updated soon |
|
|
|
## Model Card Authors |
|
Will be updated soon |
|
|
|
## Model Card Contact |
|
Will be updated soon |
|
|
