Launch of CodeGemma and Recurrent Gemma models
CodeGemma: Focus on code completion and code generation tasks, with excellent mathematical and logical reasoning capabilities
RecurrentGemma: An efficient architecture optimized for research experiments that uses recurrent neural networks and local attention to improve memory efficiency. big
Reduce memory usage, improve throughput and promote research and innovation.
Google also updated its open source Gemma model and released Gemma 1.1.
CodeGemma is a series of lightweight, state-of-the-art open models built using the same research and technology as the Gemini model was created.
The CodeGemma model uses the same architecture as the Gemma model family and is trained on more than 500 billion major code tags. As a result, the CodeGemma model achieves state-of-the-art code performance in completing and generating tasks while maintaining strong understanding and reasoning capabilities at scale.
Setup: To complete this tutorial, you first need to complete the setup instructions in Gemma Setup. Gemma setup instructions show you how to perform the following
Action: Visit Gemma at kaggle.com Select a Colab runtime with sufficient resources to run the Gemma 2B model.
Generate and configure Kaggle username and API key. After you have finished setting up Gemma, please continue to the next section, where you will set up environment variables for the Colab environment.
Select runtime library
To complete this tutorial, you need to have a Colab runtime with sufficient resources to run the CodeGemma 2B model. In this case, you can use a T4 GPU:
In the upper right corner of the Colab window, select (Other connection option).
Select Change Runtime Type.
Under Hardware Accelerator, select T4 GPU.
Configure your API key
To use Gemma, you must provide your Kaggle username and Kaggle API key.
To generate a Kaggle API key, go to the Account tab of the Kaggle user profile and select Create New Token. This will trigger the download of the kaggle.json file containing your API credentials.
In Colab, select Secrets in the left pane, and add your Kaggle username and Kaggle API key. Store your username under the name KAGGLE_USERNAME and store your API key under the name KAGGLE_KEY.
This tutorial guides you through various coding tasks using CodeGemma. To learn more about CodeGemma:
For technical specifications for CodeGemma models, please refer to the CodeGemma model card.
Learn more about how to use CodeGemma in VertexAI here.
Check out the Keras CodeGemma quick start.
If you like to learn more about the Gemma open source family, you can click on the link below the video.
Thank you for watching this video. If you like it, please subscribe and like it. thank
Technical report:https://storage.googleapis.com/deepmind-media/gemma/codegemma_report.pdf
Quick Start:https://ai.google.dev/gemma/docs/codegemma/keras_quickstart
Github：https://github.com/huggingface/llm-vscode
Video: