Gemini Nano Safetensors

Model Details

Model Name: Gemini Nano (GEMINI_XS)
Version: Gemini Nano 2 (from Chrome Canary version 2024.06.05.2205)
Type: Language Model
Format: Safetensors (converted from tensorflow)
Original Creator: Google

Model Description

This repository contains the weights of Google's Gemini Nano model converted to Safetensors format. This is an unofficial conversion not affiliated with or endorsed by Google.

Conversion Process

Converted Format: Safetensors
Conversion Time: ~7 hours and 20 minutes
Hardware Used: 32GB RAM machine
Operating System: Ubuntu Mantic Minotaur
Memory Usage: ~83% of 32GB (about 26.56GB)

Known Information

This conversion includes only the model weights, not the tokenizer, inference code, or other components needed for full functionality.

Limitations

The model's full capabilities and limitations are not known due to lack of official documentation.
This conversion does not include necessary components for immediate use (e.g., model code, inference code).

Additional Information

An 'adapter' for the model exists in the community, but its purpose and functionality are not fully understood yet.

Disclaimer: This model card is based on community efforts to convert and understand the Gemini Nano model. Information may be incomplete or subject to change as new discoveries are made.