Model Type | |
Use Cases |
Areas: | |
Applications: | Assistant-like chat, Natural language generation tasks |
|
Primary Use Cases: | Dialogue, General text generation |
|
Limitations: | Use is recommended only in English, Users must comply with license agreements and acceptable use policies. |
|
Considerations: | Developers should fine-tune models for specific languages and applications if needed. |
|
|
Supported Languages | language (English), proficiency (High) |
|
Training Details |
Data Sources: | A new mix of 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: | Llama 3 is an auto-regressive language model with an optimized transformer architecture. |
|
|
Safety Evaluation |
Methodologies: | Red teaming exercises, adversarial evaluations |
|
Risk Categories: | CBRNE risks, cybersecurity risks, child safety risks |
|
Ethical Considerations: | Safety evaluation included an assessment of the model's capability to handle sensitive topics responsibly. |
|
|
Responsible Ai Considerations |
Fairness: | Considerations related to fairness and bias addressed. |
|
Transparency: | Open approach and documentation on safety standards. |
|
Accountability: | Meta is responsible for addressing misuse that arises from the model's capabilities. |
|
Mitigation Strategies: | Safety tools such as Meta Llama Guard 2 and Code Shield are available to mitigate risks. |
|
|
Input Output |
Input Format: | |
Accepted Modalities: | |
Output Format: | |
|
Release Notes |
Version: | |
Date: | |
Notes: | Initial release of Llama 3 family models. |
|
|
|