| Model Type | | text-to-text, decoder-only, large language model |
|
| Use Cases |
| Areas: | | Content Creation and Communication, Research and Education |
|
| Applications: | | Text Generation, Chatbots and Conversational AI, Text Summarization, NLP Research, Language Learning Tools, Knowledge Exploration |
|
| Primary Use Cases: | | Generate creative text formats, Power conversational interfaces, Generate concise summaries |
|
| Limitations: | | Training Data Influences, Context and Task Complexity, Language Ambiguity and Nuance, Factual Accuracy Limitations, Common Sense Limitations |
|
| Considerations: | | LLMs are better at tasks that can be framed with clear prompts and instructions. Factual accuracy should be verified as LLMs are not knowledge bases. |
|
|
| Additional Notes | | Gemma models are designed for responsible AI development. They include open weights for pre-trained and instruction-tuned variants, enabling wide accessibility and innovation. |
|
| Supported Languages | |
| Training Details |
| Data Sources: | | Web Documents, Code, Mathematics |
|
| Data Volume: | | 8 trillion tokens for 9B model |
|
| Hardware Used: | |
| Model Architecture: | | Gemma models are built from the same research and technology used to create the Gemini models. |
|
|
| Safety Evaluation |
| Methodologies: | | Red-teaming, Benchmark testing |
|
| Risk Categories: | | Text-to-Text Content Safety, Text-to-Text Representational Harms, Memorization, Large-scale harm |
|
| Ethical Considerations: | | Models were evaluated for child safety, content safety, representational harms, memorization, large-scale harms. |
|
|
| Responsible Ai Considerations |
| Fairness: | | Models underwent careful scrutiny, input data pre-processing, and posterior evaluations to address socio-cultural biases. |
|
| Transparency: | | Details on the models' architecture, capabilities, limitations, and evaluation processes are summarized in the model card. |
|
| Accountability: | | Responsible use guidelines are provided, see the Responsible Generative AI Toolkit. |
|
| Mitigation Strategies: | | Continuous monitoring and exploration of de-biasing techniques during model training, fine-tuning, and other use cases are encouraged for mitigating perpetuation of biases. |
|
|
| Input Output |
| Input Format: | | Text string, such as a question, a prompt, or a document to be summarized. |
|
| Accepted Modalities: | |
| Output Format: | | Generated English-language text in response to the input. |
|
| Performance Tips: | | Models perform better with clear prompts and sufficient context. |
|
|