| Model Type | | text generation, instruction tuned |
|
| Use Cases |
| Areas: | |
| Applications: | | assistant-like chat, natural language generation tasks |
|
| Primary Use Cases: | |
| Limitations: | | out-of-scope for compliance violations, use in languages other than English with limitations |
|
|
| Additional Notes | | Llama 3 addresses users across different backgrounds without unnecessary judgment or normativity. |
|
| Supported Languages | | English (intended for commercial and research use) |
|
| 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 |
|
| Risk Categories: | |
| Ethical Considerations: | | Developers should perform safety testing and tuning tailored to their specific applications. |
|
|
| Responsible Ai Considerations |
| Fairness: | | openness, inclusivity and helpfulness |
|
| Transparency: | | Open approach for better and safer products |
|
| Accountability: | |
| Mitigation Strategies: | | Llama Guard and Code Shield safeguards |
|
|
| Input Output |
| Input Format: | |
| Accepted Modalities: | |
| Output Format: | |
|
| Release Notes |
| Version: | |
| Date: | |
| Notes: | | The tuned versions use SFT and RLHF to align with human preferences for helpfulness and safety. |
|
|
|