Vision Token Calculator

A Python tool for calculating the number of tokens generated when processing images with various Vision Language Models (VLMs).

Features

• Calculate image tokens for different VLMs
• Support for both existing images and dummy images
• Detailed token analysis including image size and token count
• Easy-to-use command line interface

Installation

Option 1: Install from PyPI (recommended)

pip install vt-calc

Option 2: Install as editable package (for development)

pip install -e .

Option 3: Install dependencies only

pip install -r requirements.txt

Usage

Method 1: Using the vt-calc command (after pip install -e .)

After installing with pip install -e ., you can use the vt-calc command directly:

# Using an existing image
vt-calc --image path/to/your/image.jpg

# Creating a dummy image with specific dimensions
vt-calc --size 1920 1080

# Specifying a different model
vt-calc --image path/to/your/image.jpg --model-path "model/path"

Method 2: Direct python execution

# Using an existing image
python calculate.py --image path/to/your/image.jpg

# Creating a dummy image with specific dimensions
python calculate.py --size 1920 1080

# Specifying a different model
python calculate.py --image path/to/your/image.jpg --model-path "model/path"

Supported Models

Model	Model size
Qwen2.5-VL	3B / 7B / 32B / 72B
Gemma3	4B / 12B / 27B
InternVL3	1B / 2B / 8B / 14B / 38B / 78B

License

This project is licensed under the MIT License - see the LICENSE file for details.