Model Type | text generation, auto-regressive |
|
Use Cases |
Areas: | |
Applications: | Chat assistants, Various NLP tasks |
|
Primary Use Cases: | |
Limitations: | English only, Potential for biased responses |
|
Considerations: | Adherence to the Acceptable Use Policy and suitability testing for specific use cases required. |
|
|
Additional Notes | |
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: | |
Model Architecture: | Optimized transformer architecture |
|
|
Safety Evaluation |
Methodologies: | Red teaming, Adversarial evaluations |
|
Findings: | Model is less likely to falsely refuse prompts than previous models |
|
Risk Categories: | Misinformation, Bias, Child Safety |
|
Ethical Considerations: | Residual risks will likely remain; developers should assess these risks in context of their application. |
|
|
Responsible Ai Considerations |
Fairness: | Enhancements to reduce refusals while maintaining usefulness. |
|
Transparency: | Open sourced tools for community use and contributions. |
|
Accountability: | Developers need to ensure sufficient model safety for their applications. |
|
Mitigation Strategies: | Use of Responsible Use Guide, Llama Guard, and community contributions to mitigate risks. |
|
|
Input Output |
Input Format: | Text prompts formatted as chat messages. |
|
Accepted Modalities: | |
Output Format: | Generated text or code output. |
|
|
Release Notes |
Version: | |
Date: | |
Notes: | Introduction of Meta-Llama-3-8B-Instruct with improved safety and helper functions. |
|
|
|