AI/ML Inference Library
3.5k 2026-04-18
kuprel/min-dalle
A fast, minimal PyTorch port of DALL·E Mini, optimized for efficient text-to-image generation inference.
Core Features
Fast and minimal DALL·E Mini port with mega weights.
PyTorch-based inference with minimal third-party dependencies.
High performance on various GPUs (e.g., 15 sec for 3x3 grid on A10G).
Configurable image generation parameters (seed, grid size, temperature, supercondition factor).
Supports progressive image output stream for interactive use.
Command-line interface for easy image generation.
Quick Start
pip install min-dalleDetailed Introduction
min(DALL·E) is an optimized, stripped-down PyTorch port of Boris Dayma's DALL·E Mini (with mega weights). It focuses on fast inference for text-to-image generation, requiring only numpy, requests, pillow, and torch as dependencies. The project enables users to quickly generate high-quality images from text prompts, offering configurable parameters for fine-tuning results. Its efficiency makes it suitable for local deployment, interactive applications, and integration into other AI projects, providing a streamlined alternative for DALL·E Mini users.