This article needs additional or more specific categories. Please help out by adding categories to it so that it can be listed with similar articles. (February 2023)

GPT-J

Logo
Developer(s)	EleutherAI
Initial release	June 9, 2021; 2 years ago (2021-06-09)
Type	Large language model Generative pre-trained transformer Foundation model
License	Open-source
Website	6b.eleuther.ai

GPT-J or GPT-J-6B is an open-source large language model (LLM) developed by EleutherAI in 2021. ^[1] As the name suggests, it is a generative pre-trained transformer model designed to produce human-like text that continues from a prompt. The optional "6B" in the name refers to the fact that it has 6 billion parameters. ^[2]

Architecture

GPT-J is a GPT-3-like model with 6 billion parameters. ^[3] Like GPT-3, it is an autoregressive, decoder-only transformer model designed to solve natural language processing (NLP) tasks by predicting how a piece of text will continue. ^[1]

Its architecture differs from GPT-3 in three main ways. ^[1]

• The attention and feedforward neural network were computed in parallel during training, allowing for greater efficiency.
• The GPT-J model uses rotary position embeddings, which has been found to be a superior method of injecting positional information into transformers. ^{[4] [5]}
• GPT-J uses dense attention instead of efficient sparse attention, as used in GPT-3.

Beyond that, the model has 28 transformer layers and 16 attention heads. Its vocabulary size is 50257 tokens, the same size as GPT-2's. ^[2] It has a context window size of 2048 tokens. ^[6]

It was trained on the Pile dataset, ^{[2] [3]} using the Mesh Transformer JAX library in JAX to handle the parallelization scheme. ^{[2] [7]}

Performance

GPT-J was designed to generate English text from a prompt. It was not designed for translating or generating text in other languages or for performance without first fine-tuning the model for a specific task. ^[2] Nonetheless, GPT-J performs reasonably well even without fine-tuning, even in translation (at least from English to French). ^[8]

When neither is fine-tuned, GPT-J-6B performs almost as well as the 6.7 billion parameter GPT-3 (Curie) on a variety of tasks. ^[3] It even outperforms the 175 billion parameter GPT-3 (Davinci) on code generation tasks. ^{[9] [10]} With fine-tuning, it outperforms an untuned GPT-3 (Davinci) on a number of tasks. ^[1]

Like all LLMs, it is not programmed to give factually accurate information, only to generate text based on probability. ^[2]

Applications

The untuned GPT-J is available on EleutherAI's website, ^[11] NVIDIA's Triton Inference Server, ^[12] and NLP Cloud's website. ^[13] Cerebras ^[1] and Amazon Web Services ^{[14] [15]} offer services to fine-tune the GPT-J model for company-specific tasks. Graphcore offers both fine-tuning and hosting services for the untuned GPT-J, as well as offering to host the fine-tuned models after they are produced. ^[16] CoreWeave offers hosting services for both the untuned GPT-J and fine-tuned variants. ^{[17] [18]}

In March 2023, Databricks released Dolly, an Apache-licensed, instruction-following model created by fine-tuning GPT-J on the Stanford Alpaca dataset. ^[19] NovelAI's Sigurd ^[20] and Genji-JP 6B ^[21] models are both fine-tuned versions of GPT-J. They also offer further fine-tuning services to produce and host custom models. ^[22]

EleutherAI has received praise from Cerebras, ^[1] GPT-3 Demo, ^[3] NLP Cloud, ^[13] and Databricks ^[19] for making the model open-source, and its open-source status is often cited as a major advantage when choosing which model to use. ^{[10] [16] [23]}

References