| Model Type | | large language models, text generation |
|
| Use Cases |
| Areas: | |
| Applications: | | assistant-like chat, natural language generation |
|
| Limitations: | | Use in languages other than English |
|
| Considerations: | | Developers must follow the Community License and Acceptable Use Policy. |
|
|
| Additional Notes | | This is a static model trained on an offline dataset. |
|
| Training Details |
| Data Sources: | | publicly available online data |
|
| Data Volume: | |
| Methodology: | | supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) |
|
| Context Length: | |
| Hardware Used: | | Meta's Research SuperCluster, third-party cloud compute |
|
| Model Architecture: | | optimized transformer architecture |
|
|
| Safety Evaluation |
| Methodologies: | | red teaming, adversarial evaluations |
|
| Findings: | | Lower residual risks compared to previous versions, Emphasis on avoiding false refusals |
|
| Risk Categories: | | Cybersecurity, Child Safety |
|
| Ethical Considerations: | | Focus on reducing exposure to harmful or malicious outputs. |
|
|
| Responsible Ai Considerations |
| Mitigation Strategies: | | Use of tools like Meta Llama Guard 2 and Code Shield to reduce risks. |
|
|
| Input Output |
| Input Format: | |
| Output Format: | |
|
| Release Notes |
| Version: | |
| Date: | |
| Notes: | | Instruction tuned and optimized for dialogue use cases. |
|
|
|