| Model Type | | text generation, causal-lm |
|
| Use Cases |
| Areas: | |
| Applications: | | Development of chat assistants, Sentiment analysis, Summarization |
|
| Primary Use Cases: | | Natural language understanding and generation tasks |
|
| Limitations: | | Limited to Arabic and English, Not suitable for high-stakes decisions |
|
| Considerations: | | Not to rely solely on model's outputs. |
|
|
| Additional Notes | | The strategies for pre-training, fine-tuning and adaptation to Arabic are extensible to other low and medium resource languages. |
|
| Supported Languages | | Arabic (High proficiency), English (Strong proficiency) |
|
| Training Details |
| Data Sources: | | Web, Code, Books, Scientific, Synthetic |
|
| Data Volume: | |
| Methodology: | | SwiGLU non-linear activation, ALiBi position encoding |
|
| Context Length: | |
| Hardware Used: | | 64 Cerebras CS-2 Wafer-Scale Engines (WSE-2) |
|
| Model Architecture: | | Transformer-based, decoder-only architecture |
|
|
| Safety Evaluation |
| Methodologies: | | Various dimensions including knowledge, reasoning, misinformation/bias |
|
| Risk Categories: | |
| Ethical Considerations: | | Understanding limitations and potential misuse. |
|
|
| Responsible Ai Considerations |
| Mitigation Strategies: | | Various techniques to reduce bias. |
|
|
| Input Output |
| Input Format: | |
| Accepted Modalities: | |
| Output Format: | |
|
| Release Notes |
| Version: | |
| Notes: | | Release of the Jais family with improved cultural understanding for Arabic. |
|
|
|
