tensorkiko 0.1.8

A fast and intuitive tool for visualizing and analyzing model structures from safetensors files

TensorKiko

A fast and intuitive tool for visualizing and analyzing model structures from safetensors files, supporting tree-based visualizations and detailed parameter analysis.

Installation

Requirements

• Python 3.11 or higher

Installation Methods

Using pip

To install TensorKiko with pip, run the following command:

pip install tensorkiko

Using Homebrew

To install TensorKiko using Homebrew, use the following commands:

brew tap takara-ai/tensorkiko https://github.com/takara-ai/TensorKiko
brew install tensorkiko

Usage

After installation, you can use TensorKiko from the command line to visualize your model:

tensorkiko path/to/your/model.safetensors

TensorKiko can also attempt to convert .ckpt files to .safetensors. However, conversion will only succeed if the .ckpt file is in a standard model format without unique code elements such as SQLite.

For more options, use the help command:

tensorkiko --help

Features

• Load and process safetensors files
• Generate interactive HTML visualizations of model structures
• Analyze model parameters, memory usage, and estimated FLOPs
• Search functionality for easy navigation of large models

Contributing

We welcome contributions! If you'd like to contribute, please feel free to submit a Pull Request.

License

This project is licensed under the MIT License.