1kxia/gemma-3-270m
Gemma 3 270M is a 0.3 billion parameter multimodal model from Google DeepMind, built from the same research as Gemini models. It handles text and image input, generating text output, and supports a 32K token context window. This lightweight model is designed for text generation and image understanding tasks, including question answering and summarization, and is suitable for resource-limited environments.
Loading preview...
Gemma 3 270M Overview
Gemma 3 270M is part of Google DeepMind's Gemma family, a series of lightweight, open multimodal models derived from the Gemini research. This 0.3 billion parameter model is capable of processing both text and image inputs (images normalized to 896x896 resolution, encoded to 256 tokens) and generating text outputs. It features a 32K token context window and supports over 140 languages.
Key Capabilities
- Multimodal Understanding: Processes text and images for tasks like image analysis and visual data extraction.
- Text Generation: Excels in creative text formats, summarization, and conversational AI.
- Multilingual Support: Trained on web documents in over 140 languages.
- Resource-Efficient: Its compact size allows deployment in environments with limited computational resources, such as laptops and desktops.
Training and Performance
The model was trained on 6 trillion tokens, including diverse web documents, code, and mathematical texts, with a knowledge cutoff of August 2024. It demonstrates competitive performance on various benchmarks for its size, including HellaSwag (40.9 PT, 37.7 IT), PIQA (67.7 PT, 66.2 IT), and WinoGrande (52.0 PT, 52.3 IT).
Intended Use Cases
- Content Creation: Generating creative text, marketing copy, and email drafts.
- Conversational AI: Powering chatbots and virtual assistants.
- Research & Education: Serving as a foundation for VLM/NLP research and language learning tools.