| Model Type | | text generation, large language model |
|
| Use Cases |
| Areas: | | commercial use, research, assistant-like chat |
|
| Applications: | | natural language generation tasks |
|
| Primary Use Cases: | | dialogue models, instruction tuned models |
|
| Limitations: | | English language only, no use outside acceptable use policy |
|
| Considerations: | | Developers may fine-tune for languages beyond English compliant with license terms |
|
|
| Additional Notes | | Model optimized for dialogue use cases with community feedback |
|
| Supported Languages | |
| Training Details |
| Data Sources: | | publicly available online data |
|
| Data Volume: | |
| Methodology: | | supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) |
|
| Context Length: | |
| Training Time: | |
| Hardware Used: | | Meta's Research SuperCluster, third-party cloud compute |
|
| Model Architecture: | | optimized transformer architecture |
|
|
| Safety Evaluation |
| Methodologies: | | internal evaluations, red teaming, adversarial evaluations |
|
| Findings: | | improved fine-tuning to limit false refusals |
|
| Risk Categories: | | misuse, CBRNE threats, cybersecurity dangers |
|
| Ethical Considerations: | | Responsible AI development with community feedback |
|
|
| Responsible Ai Considerations |
| Fairness: | | Designed to serve a wide range of users with no normativity |
|
| Transparency: | | Open approach to community feedback and development |
|
| Accountability: | | Developers responsible for deployment use-case safety |
|
| Mitigation Strategies: | | Llama Guard, Purple Llama |
|
|
| Input Output |
| Input Format: | |
| Accepted Modalities: | |
| Output Format: | |
|
| Release Notes |
| Version: | |
| Date: | |
| Notes: | | Release of Llama 3 models in sizes 8B and 70B with instruction tuning |
|
|
|