gemini

Introduction

Gemini is a transformer model based on Google's T5 architecture, designed for summarizing and explaining code snippets in multiple programming languages. It is pre-trained on a substantial dataset and fine-tuned to generate higher-level explanations.

Architecture

Gemini leverages the T5 architecture and is pre-trained on 800k code/description pairs. It is fine-tuned on 10k synthetically generated higher-level explanations. The model supports languages such as Python, JavaScript (including frameworks like React), Java, Ruby, and Go, outputting descriptions in English.

Training

• Pre-training: 800k code/description pairs.
• Fine-tuning: 10k synthetically generated higher-level code explanations.
• Model Sizes:
  • Gemini: 770 million parameters.
  • Gemini-Small: 220 million parameters.

Guide: Running Locally

To run Gemini locally:

1. Install Transformers: Make sure you have the transformers library installed.
2. Import and Setup:

   from transformers import pipeline
   
   summarizer = pipeline('text2text-generation', model='describeai/gemini')
   

3. Run Summarization:

   code = "print('hello world!')"
   response = summarizer(code, max_length=100, num_beams=3)
   print("Summarized code: " + response[0]['generated_text'])
   

4. Suggested Environment: Use cloud GPUs for more efficient processing, such as those provided by AWS, Google Cloud, or Azure.

License

Gemini is released under the MIT License, allowing for flexible use and distribution.