Model parallel transformers in JAX and Haiku

Overview

Table of contents

  1. Mesh Transformer JAX
    1. Updates
  2. Pretrained Models
    1. GPT-J-6B
      1. Links
      2. Acknowledgments
      3. License
      4. Model Details
      5. Zero-Shot Evaluations
  3. Architecture and Usage
    1. Fine-tuning
    2. JAX Dependency
  4. TODO

Mesh Transformer JAX

A haiku library using the xmap/pjit operators in JAX for model parallelism of transformers.

The parallelism scheme is similar to the original Megatron-LM, which is efficient on TPUs due to the high speed 2d mesh network. There is also an experimental model version which implements ZeRo style sharding.

This library is designed for scalability up to approximately 40B parameters on TPUv3s, beyond which different parallelism strategies should be used. See other implementations such as GPT-NeoX or DeepSpeed for that.

One future direction for research is integrating this codebase with swarm-jax, to achieve further scalability with pipeline parallelism.

Updates

12-07-21: Added guide to fine tuning

Pretrained Models

GPT-J-6B

A 6 billion parameter, autoregressive text generation model trained on The Pile.

Links

Slim weights (bf16 weights only, for inference, 9GB)

Full weights (including optimizer params, 61GB)

Colab demo

Web demo

Aran's blog post

Acknowledgments

This project would not have been possible without compute generously provided by the TPU Research Cloud with assistance from EleutherAI.

Thanks to the Cloud TPU team at Google for providing early access to the Cloud TPU VM alpha (now publicly available!)

Thanks to everyone who have helped out one way or another (listed alphabetically):

  • Aran Komatsuzaki for advice with experiment design and writing the blog posts.
  • James Bradbury for valuable assistance with debugging JAX issues.
  • Janko Prester for creating the web demo frontend.
  • Laurence Golding for adding some features to the web demo.
  • Leo Gao for running zero shot evaluations for the baseline models for the table.

License

The weights of GPT-J-6B are licensed under version 2.0 of the Apache License.

Model Details

Hyperparameter Value
n_parameters 6,053,381,344
n_layers 28*
d_model 4,096
d_ff 16,384
n_heads 16
d_head 256
n_ctx 2,048
n_vocab 50,257 (same tokenizer as GPT-2/3)
position encoding Rotary position encodings (RoPE)
RoPE dimensions 64

* each layer consists of one feedforward block and one self attention block

The model consists of 28 layers with a model dimension of 4096, and a feedforward dimension of 16384. The model dimension is split into 16 heads, each with a dimension of 256. Rotary position encodings (RoPE) was applied to 64 dimensions of each head. The model is trained with a tokenization vocabulary of 50257, using the same set of BPEs as GPT-2/GPT-3.

Zero-Shot Evaluations

Models roughly sorted by performance, or by FLOPs if not available.

Model Weights Training FLOPs LAMBADA PPL ↓ LAMBADA Acc ↑ Winogrande ↑ Hellaswag ↑ PIQA ↑ Dataset Size (GB)
Chance 0 ~a lot ~0% 50% 25% 25% 0
GPT-3-Ada‡ ----- 9.95 51.6% 52.9% 43.4% 70.5% -----
GPT-2-1.5B ----- 10.63 51.21% 59.4% 50.9% 70.8% 40
GPTNeo-1.3B‡ 3.0e21 7.50 57.2% 55.0% 48.9% 71.1% 825
Megatron-2.5B* 2.4e21 ----- 61.7% ----- ----- ----- 174
GPTNeo-2.7B‡ 6.8e21 5.63 62.2% 56.5% 55.8% 73.0% 825
GPT-3-1.3B*‡ 2.4e21 5.44 63.6% 58.7% 54.7% 75.1% ~800
GPT-3-Babbage‡ ----- 5.58 62.4% 59.0% 54.5% 75.5% -----
Megatron-8.3B* 7.8e21 ----- 66.5% ----- ----- ----- 174
GPT-3-2.7B*‡ 4.8e21 4.60 67.1% 62.3% 62.8% 75.6% ~800
Megatron-11B† 1.0e22 ----- ----- ----- ----- ----- 161
GPT-J-6B 1.5e22 3.99 69.7% 65.3% 66.1% 76.5% 825
GPT-3-6.7B*‡ 1.2e22 4.00 70.3% 64.5% 67.4% 78.0% ~800
GPT-3-Curie‡ ----- 4.00 69.3% 65.6% 68.5% 77.9% -----
GPT-3-13B*‡ 2.3e22 3.56 72.5% 67.9% 70.9% 78.5% ~800
GPT-3-175B*‡ 3.1e23 3.00 76.2% 70.2% 78.9% 81.0% ~800
GPT-3-Davinci‡ ----- 3.0 75% 72% 78% 80% -----
Gopher 230B* 6.31E+23 ----- 74.50% 70.10% 79.20% 81.80% 1344
MT-NLG 530B*‡ ----- ----- 76.6% 73.0% 80.2% 82.0% -----

* represents evaluation numbers reported by their respective authors, all other numbers are provided by running the lm-evaluation-harness either with the released weights or with API access. Due to subtle implementation differences as well as different zero shot task framing, these might not be directly comparable. See this blog post for more details.

The Megatron-11B model provides no comparable metrics, and several implementations using the released weights do not reproduce the generation quality and evaluations. (see 1 2 3) Thus, evaluation was not attempted.

These models have been trained with data which contains possible test set contamination. The OpenAI GPT-3 models failed to deduplicate training data for certain test sets, while the GPT-Neo models as well as this one is trained on The Pile, which has not been deduplicated against any test sets.

Architecture and Usage

Most scripts in this repository are designed to be run on TPUs, which under the TPU-VM architecture are virtual machines which can run arbitrary code. Most scripts are designed to spin up a TPU, SSH into it to set up the dependencies and copy code over from the local directory, and then start a Ray worker which can accept RPC calls.

The TPUVMs handles running model training steps and evaluation, checkpoint save and loading, while the driver python program handles data loading and general orchestration (such as when to save checkpoints etc).

This means that most scripts (train.py, eval_harness.py etc) expect to be running on a GCE virtual machine in the same region as the TPUs, to minimize RPC latency and data transfer cost. Other scripts (usually ones which don't take a --tpu argument, such as device_sample.py, device_serve.py or device_train.py) expect to be run directly on a TPUVM. The device_* scripts only work on a v3-8 and not on larger pods.

Furthermore, there is an example (resharding_example.py) of how to convert the provided checkpoints (which have 8 shards in the case of GPT-J-6B) down to a smaller number, such as for when running on GPU(s).

Fine-tuning

To fine-tune the model, run device_train.py on a TPU VM. Using a TPU v3-8, you can fine-tune at a rate of ~5000 tokens/second, which should be sufficient for small-to-medium-size datasets.

Please read the step by step guide for thorough fine-tuning instructions.

JAX Dependency

Note this library has some specific requirements for JAX version. Specifically, to use the v1 models (including GPT-J 6B), jax==0.2.12 is required. This in turn depends on jaxlib==0.1.68. If this is not done, you will get cryptic xmap errors

However, to use the v2 model code (no publicly released weights), the newest JAX version can be used.

Citation

To cite this repository:

@misc{mesh-transformer-jax,
  author = {Wang, Ben},
  title = {{Mesh-Transformer-JAX: Model-Parallel Implementation of Transformer Language Model with JAX}},
  howpublished = {\url{https://github.com/kingoflolz/mesh-transformer-jax}},
  year = 2021,
  month = May
}

To cite the weights of GPT-J-6B:

@misc{gpt-j,
  author = {Wang, Ben and Komatsuzaki, Aran},
  title = {{GPT-J-6B: A 6 Billion Parameter Autoregressive Language Model}},
  howpublished = {\url{https://github.com/kingoflolz/mesh-transformer-jax}},
  year = 2021,
  month = May
}

If you use this repository or any of the pretrained weights to do something cool, we would love to hear about it. Feel free to open a github issue or reach out over email (in profile).

TODO

  • disentangle heads and shards
  • test/benchmark on TPU
  • implement gradient checkpointing
  • fix initialization
  • mixed precision
  • deal with preemptible TPUs
  • test and validate generation
  • shard activations instead of replicating for memory efficiency (in v2)
  • support ZeRO style sharding (in v2)
Owner
Ben Wang
Ben Wang
Pipeline for fast building text classification TF-IDF + LogReg baselines.

Text Classification Baseline Pipeline for fast building text classification TF-IDF + LogReg baselines. Usage Instead of writing custom code for specif

Dani El-Ayyass 57 Dec 07, 2022
TLA - Twitter Linguistic Analysis

TLA - Twitter Linguistic Analysis Tool for linguistic analysis of communities TLA is built using PyTorch, Transformers and several other State-of-the-

Tushar Sarkar 47 Aug 14, 2022
Code Implementation of "Learning Span-Level Interactions for Aspect Sentiment Triplet Extraction".

Span-ASTE: Learning Span-Level Interactions for Aspect Sentiment Triplet Extraction ***** New March 31th, 2022: Scikit-Style API for Easy Usage *****

Chia Yew Ken 111 Dec 23, 2022
BERT-based Financial Question Answering System

BERT-based Financial Question Answering System In this example, we use Jina, PyTorch, and Hugging Face transformers to build a production-ready BERT-b

Bithiah Yuan 61 Sep 18, 2022
NLP, before and after spaCy

textacy: NLP, before and after spaCy textacy is a Python library for performing a variety of natural language processing (NLP) tasks, built on the hig

Chartbeat Labs Projects 2k Jan 04, 2023
Just a basic Telegram AI chat bot written in Python using Pyrogram.

Nikko ChatBot Just a basic Telegram AI chat bot written in Python using Pyrogram. Requirements Python 3.7 or higher. A bot token. Installation $ https

ʀᴇxɪɴᴀᴢᴏʀ 2 Oct 21, 2022
Sample data associated with the Aurora-BP study

The Aurora-BP Study and Dataset This repository contains sample code, sample data, and explanatory information for working with the Aurora-BP dataset

Microsoft 16 Dec 12, 2022
An open collection of annotated voices in Japanese language

声庭 (Koniwa): オープンな日本語音声とアノテーションのコレクション Koniwa (声庭): An open collection of annotated voices in Japanese language 概要 Koniwa(声庭)は利用・修正・再配布が自由でオープンな音声とアノテ

Koniwa project 32 Dec 14, 2022
FireFlyer Record file format, writer and reader for DL training samples.

FFRecord The FFRecord format is a simple format for storing a sequence of binary records developed by HFAiLab, which supports random access and Linux

77 Jan 04, 2023
Codes for processing meeting summarization datasets AMI and ICSI.

Meeting Summarization Dataset Meeting plays an essential part in our daily life, which allows us to share information and collaborate with others. Wit

xcfeng 39 Dec 14, 2022
An assignment from my grad-level data mining course demonstrating some experience with NLP/neural networks/Pytorch

NLP-Pytorch-Assignment An assignment from my grad-level data mining course (before I started personal projects) demonstrating some experience with NLP

David Thorne 0 Feb 06, 2022
ConferencingSpeech2022; Non-intrusive Objective Speech Quality Assessment (NISQA) Challenge

ConferencingSpeech 2022 challenge This repository contains the datasets list and scripts required for the ConferencingSpeech 2022 challenge. For more

21 Dec 02, 2022
Python wrapper for Stanford CoreNLP tools v3.4.1

Python interface to Stanford Core NLP tools v3.4.1 This is a Python wrapper for Stanford University's NLP group's Java-based CoreNLP tools. It can eit

Dustin Smith 610 Sep 07, 2022
This simple Python program calculates a love score based on your and your crush's full names in English

This simple Python program calculates a love score based on your and your crush's full names in English. There is no logic or reason in the calculation behind the love score. The calculation could ha

p.katekomol 1 Jan 24, 2022
Chinese NER with albert/electra or other bert descendable model (keras)

Chinese NLP (albert/electra with Keras) Named Entity Recognization Project Structure ./ ├── NER │   ├── __init__.py │   ├── log

2 Nov 20, 2022
PyTorch implementation of convolutional neural networks-based text-to-speech synthesis models

Deepvoice3_pytorch PyTorch implementation of convolutional networks-based text-to-speech synthesis models: arXiv:1710.07654: Deep Voice 3: Scaling Tex

Ryuichi Yamamoto 1.8k Dec 30, 2022
This repository contains Python scripts for extracting linguistic features from Filipino texts.

Filipino Text Linguistic Feature Extractors This repository contains scripts for extracting linguistic features from Filipino texts. The scripts were

Joseph Imperial 1 Oct 05, 2021
A framework for evaluating Knowledge Graph Embedding Models in a fine-grained manner.

A framework for evaluating Knowledge Graph Embedding Models in a fine-grained manner.

NEC Laboratories Europe 13 Sep 08, 2022
A tool helps build a talk preview image by combining the given background image and talk event description

talk-preview-img-builder A tool helps build a talk preview image by combining the given background image and talk event description Installation and U

PyCon Taiwan 4 Aug 20, 2022