Cerebras-GPT: Open Compute-Optimal Language Models Trained on the Cerebras Wafer-Scale Cluster

TL;DR

This paper introduces Cerebras-GPT, a family of open, compute-efficient large language models scaled from 111 million to 13 billion parameters, trained on the Eleuther Pile dataset, demonstrating state-of-the-art efficiency and scalability.

Contribution

The paper presents the first open, reproducible comparison of compute-optimal scaling for large language models using the Cerebras Wafer-Scale Cluster, incorporating Maximal Update Parameterization for improved performance.

Findings

Cerebras-GPT models achieve state-of-the-art training efficiency.

Power-law scaling of model performance is predictable.

Maximal Update Parameterization enhances large model scaling.

Abstract

We study recent research advances that improve large language models through efficient pre-training and scaling, and open datasets and tools. We combine these advances to introduce Cerebras-GPT, a family of open compute-optimal language models scaled from 111M to 13B parameters. We train Cerebras-GPT models on the Eleuther Pile dataset following DeepMind Chinchilla scaling rules for efficient pre-training (highest accuracy for a given compute budget). We characterize the predictable power-law scaling and compare Cerebras-GPT with other publicly-available models to show all Cerebras-GPT models have state-of-the-art training efficiency on both pre-training and downstream objectives. We describe our learnings including how Maximal Update Parameterization ($\mu$P) can further improve large model scaling, improving accuracy and hyperparameter predictability at scale. We release our pre-trained models and code, making this paper the first open and reproducible work comparing compute-optimal model scaling to models trained on fixed dataset sizes. Cerebras-GPT models are available on HuggingFace: https://huggingface.co/cerebras.