picotron  by huggingface

Minimalist distributed training framework for educational use

Created 1 year ago
1,799 stars

Top 23.9% on SourcePulse

GitHubView on GitHub
Project Summary

Picotron is a minimalist, hackable distributed training framework designed for educational purposes, enabling users to learn and experiment with 4D parallelism (Data, Tensor, Pipeline, Context) for large language models. It offers a simplified codebase, making complex distributed training concepts accessible to researchers and students.

How It Works

Picotron implements 4D parallelism by breaking down model training across data, tensor, pipeline, and context dimensions. This approach allows for efficient distribution of large models and datasets across multiple GPUs, facilitating training of models that would otherwise be too large for single devices. The framework prioritizes code readability and simplicity, with core components like train.py and parallelism strategies under 300 lines each.

Quick Start & Requirements

Health Check
Last Commit

3 weeks ago

Responsiveness

1 week

Pull Requests (30d)
1
Issues (30d)
0
Star History
129 stars in the last 30 days

Explore Similar Projects

Starred by Jiaming Song Jiaming Song(Chief Scientist at Luma AI), Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), and
6 more.

LLaMA-Adapter by OpenGVLab

0.1%
6k
Efficient fine-tuning for instruction-following LLaMA models
Created 2 years ago
Updated 1 year ago
Starred by Andrej Karpathy Andrej Karpathy(Founder of Eureka Labs; Formerly at Tesla, OpenAI; Author of CS 231n), George Hotz George Hotz(Author of tinygrad; Founder of the tiny corp, comma.ai), and
20 more.

TinyLlama by jzhang38

0.1%
9k
Tiny pretraining project for a 1.1B Llama model
Created 2 years ago
Updated 1 year ago
Starred by Tobi Lutke Tobi Lutke(Cofounder of Shopify), Li Jiang Li Jiang(Coauthor of AutoGen; Engineer at Microsoft), and
26 more.

ColossalAI by hpcaitech

0.1%
41k
AI system for large-scale parallel training
Created 3 years ago
Updated 13 hours ago
Feedback? Help us improve.