Training BERT with Compute/Time (Academic) Budget

Overview

Training BERT with Compute/Time (Academic) Budget

This repository contains scripts for pre-training and finetuning BERT-like models with limited time and compute budget. The code is based on the work presented in the following paper:

Peter Izsak, Moshe Berchansky, Omer Levy, How to Train BERT with an Academic Budget - (to appear at EMNLP 2021).

Installation

The pre-training and finetuning scripts are based on Deepspeed and HuggingFace Transformers libraries.

Preliminary Installation

We recommend creating a virtual environment with python 3.6+, PyTorch and apex.

Installation Requirements

pip install -r requirements.txt

We suggest running Deepspeed's utility ds_report and verify Deepspeed components can be compiled (JIT).

Dataset

The dataset directory includes scripts to pre-process the datasets we used in our experiments (Wikipedia, Bookcorpus). See dedicated README for full details.

Pretraining

Pretraining script: run_pretraining.py

For all possible pretraining arguments see: python run_pretraining.py -h

We highly suggest reviewing the various training features we provide within the library.

Example for training with the best configuration presented in our paper (24-layers/1024H/time-based learning rate schedule/fp16):
deepspeed run_pretraining.py \
  --model_type bert-mlm --tokenizer_name bert-large-uncased \
  --hidden_act gelu \
  --hidden_size 1024 \
  --num_hidden_layers 24 \
  --num_attention_heads 16 \
  --intermediate_size 4096 \
  --hidden_dropout_prob 0.1 \
  --attention_probs_dropout_prob 0.1 \
  --encoder_ln_mode pre-ln \
  --lr 1e-3 \
  --train_batch_size 4096 \
  --train_micro_batch_size_per_gpu 32 \
  --lr_schedule time \
  --curve linear \
  --warmup_proportion 0.06 \
  --gradient_clipping 0.0 \
  --optimizer_type adamw \
  --weight_decay 0.01 \
  --adam_beta1 0.9 \
  --adam_beta2 0.98 \
  --adam_eps 1e-6 \
  --total_training_time 24.0 \
  --early_exit_time_marker 24.0 \
  --dataset_path <dataset path> \
  --output_dir /tmp/training-out \
  --print_steps 100 \
  --num_epochs_between_checkpoints 10000 \
  --job_name pretraining_experiment \
  --project_name budget-bert-pretraining \
  --validation_epochs 3 \
  --validation_epochs_begin 1 \
  --validation_epochs_end 1 \
  --validation_begin_proportion 0.05 \
  --validation_end_proportion 0.01 \
  --validation_micro_batch 16 \
  --deepspeed \
  --data_loader_type dist \
  --do_validation \
  --use_early_stopping \
  --early_stop_time 180 \
  --early_stop_eval_loss 6 \
  --seed 42 \
  --fp16

Time-based Training

Pretraining can be limited to a time-based value by defining --total_training_time=24.0 (24 hours for example).

Time-based Learning Rate Scheduling

The learning rate can be scheduled to change according to the configured total training time. The argument --total_training_time controls the total time assigned for the trainer to run, and must be specified in order to use time-based learning rate scheduling.

Time-based Learning rate schedule

To select time-based learning rate scheduling, define --lr_schedule time, and define a shape for for the annealing curve (--curve=linear for example, as seen in the figure). The warmup phase of the learning rate is define by specifying a proportion (--warmup_proportion) which accounts for the time-budget proportion available in the training session (as defined by --total_training_time). For example, for a 24 hour training session, warmup_proportion=0.1 would account for 10% of 24 hours, that is, 2.4 hours (or 144 minutes) to reach peak learning rate. The learning rate will then be scheduled to reach 0 at the end of the time budget. We refer to the provided figure for an example.

Checkpoints and Finetune Checkpoints

There are 2 types of checkpoints that can be enabled:

  • Training checkpoint - saves model weights, optimizer state and training args. Defined by --num_epochs_between_checkpoints.
  • Finetuning checkpoint - saves model weights and configuration to be used for finetuning later on. Defined by --finetune_time_markers.

finetune_time_markers can be assigned multiple points in the training time-budget by providing a list of time markers of the overall training progress. For example --finetune_time_markers=0.5 will save a finetuning checkpoint when reaching 50% of training time budget. For multiple finetuning checkpoints, use commas without space 0.5,0.6,0.9.

Validation Scheduling

Enable validation while pre-training with --do_validation

Control the number of epochs between validation runs with --validation_epochs=

To control the amount of validation runs in the beginning and end (running more that validation_epochs) use validation_begin_proportion and validation_end_proportion to specify the proportion of time and, validation_epochs_begin and validation_epochs_end to control the custom values accordingly.

Mixed Precision Training

Mixed precision is supported by adding --fp16. Use --fp16_backend=ds to use Deepspeed's mixed precision backend and --fp16_backend=apex for apex (--fp16_opt controls optimization level).

Finetuning

Use run_glue.py to run finetuning for a saved checkpoint on GLUE tasks.

The finetuning script is identical to the one provided by Huggingface with the addition of our model.

For all possible pretraining arguments see: python run_glue.py -h

Example for finetuning on MRPC:
python run_glue.py \
  --model_name_or_path <path to model> \
  --task_name MRPC \
  --max_seq_length 128 \
  --output_dir /tmp/finetuning \
  --overwrite_output_dir \
  --do_train --do_eval \
  --evaluation_strategy steps \
  --per_device_train_batch_size 32 --gradient_accumulation_steps 1 \
  --per_device_eval_batch_size 32 \
  --learning_rate 5e-5 \
  --weight_decay 0.01 \
  --eval_steps 50 --evaluation_strategy steps \
  --max_grad_norm 1.0 \
  --num_train_epochs 5 \
  --lr_scheduler_type polynomial \
  --warmup_steps 50

Generating Pretraining Commands

We provide a useful script for generating multiple (or single) pretraining commands by using python generate_training_commands.py.

python generate_training_commands.py -h

	--param_file PARAM_FILE Hyperparameter and configuration yaml
  	--job_name JOB_NAME   job name
 	--init_cmd INIT_CMD   initialization command (deepspeed or python directly)

A parameter yaml must be defined with 2 main keys: hyperparameters with argument values defined as a list of possible values, and default_parameters as default values. Each generated command will be a possible combination of the various arguments specified in the hyperparameters section.

Example:

hyperparameters:
  param1: [val1, val2]
  param2: [val1, val2]

default_parameters:
  param3: 0.0

will result in:

deepspeed run_pretraining.py --param1=val1 --param2=val1 --param3=0.0
deepspeed run_pretraining.py --param1=val1 --param2=val2 --param3=0.0
deepspeed run_pretraining.py --param1=val2 --param2=val1 --param3=0.0
deepspeed run_pretraining.py --param1=val2 --param2=val2 --param3=0.0

Citation

If you find this paper or this code useful, please cite this paper:

@article{izsak2021,
  author={Izsak, Peter and Berchansky, Moshe and Levy, Omer},
  title={How to Train BERT with an Academic Budget},
  journal={arXiv preprint arXiv:2104.07705},
  url = {https://arxiv.org/abs/2104.07705} 
  year={2021}
}
Owner
Intel Labs
Intel Labs
Using Tensorflow Object Detection API to detect Waymo open dataset

Waymo-2D-Object-Detection Using Tensorflow Object Detection API to detect Waymo open dataset Result CenterNet Training Loss SSD ResNet Training Loss C

76 Dec 12, 2022
[NeurIPS 2021 Spotlight] Code for Learning to Compose Visual Relations

Learning to Compose Visual Relations This is the pytorch codebase for the NeurIPS 2021 Spotlight paper Learning to Compose Visual Relations. Demo Imag

Nan Liu 88 Jan 04, 2023
A large-scale database for graph representation learning

A large-scale database for graph representation learning

Scott Freitas 29 Nov 25, 2022
Finetuning Pipeline

KLUE Baseline Korean(한국어) KLUE-baseline contains the baseline code for the Korean Language Understanding Evaluation (KLUE) benchmark. See our paper fo

74 Dec 13, 2022
[ICML'21] Estimate the accuracy of the classifier in various environments through self-supervision

What Does Rotation Prediction Tell Us about Classifier Accuracy under Varying Testing Environments? [Paper] [ICML'21 Project] PyTorch Implementation T

24 Oct 26, 2022
QICK: Quantum Instrumentation Control Kit

QICK: Quantum Instrumentation Control Kit The QICK is a kit of firmware and software to use the Xilinx RFSoC to control quantum systems. It consists o

81 Dec 15, 2022
OpenMMLab Detection Toolbox and Benchmark

MMDetection is an open source object detection toolbox based on PyTorch. It is a part of the OpenMMLab project.

OpenMMLab 22.5k Jan 05, 2023
How to Become More Salient? Surfacing Representation Biases of the Saliency Prediction Model

How to Become More Salient? Surfacing Representation Biases of the Saliency Prediction Model

Bogdan Kulynych 49 Nov 05, 2022
Talk covering the features of skorch

Skorch Talk Skorch - A Union of Scikit-learn and PyTorch Presentation The slides can be downloaded at: download link. Google Colab Part One - MNIST Pa

Thomas J. Fan 3 Oct 20, 2020
Tutorial page of the Climate Hack, the greatest hackathon ever

Tutorial page of the Climate Hack, the greatest hackathon ever

UCL Artificial Intelligence Society 12 Jul 02, 2022
[NeurIPS 2021] Galerkin Transformer: a linear attention without softmax

[NeurIPS 2021] Galerkin Transformer: linear attention without softmax Summary A non-numerical analyst oriented explanation on Toward Data Science abou

Shuhao Cao 159 Dec 20, 2022
Seach Losses of our paper 'Loss Function Discovery for Object Detection via Convergence-Simulation Driven Search', accepted by ICLR 2021.

CSE-Autoloss Designing proper loss functions for vision tasks has been a long-standing research direction to advance the capability of existing models

Peidong Liu(刘沛东) 54 Dec 17, 2022
Detecting Potentially Harmful and Protective Suicide-related Content on Twitter

TwitterSuicideML Scripts for reproducing the Machine Learning analysis of the paper: Detecting Potentially Harmful and Protective Suicide-related Cont

3 Oct 17, 2022
Official code for the paper "Self-Supervised Prototypical Transfer Learning for Few-Shot Classification"

Self-Supervised Prototypical Transfer Learning for Few-Shot Classification This repository contains the reference source code and pre-trained models (

EPFL INDY 44 Nov 04, 2022
This project deploys a yolo fastest model in the form of tflite on raspberry 3b+. The model is from another repository of mine called -Trash-Classification-Car

Deploy-yolo-fastest-tflite-on-raspberry 觉得有用的话可以顺手点个star嗷 这个项目将垃圾分类小车中的tflite模型移植到了树莓派3b+上面。 该项目主要是为了记录在树莓派部署yolo fastest tflite的流程 (之后有时间会尝试用C++部署来提升

7 Aug 16, 2022
Code repository for the paper "Doubly-Trained Adversarial Data Augmentation for Neural Machine Translation" with instructions to reproduce the results.

Doubly Trained Neural Machine Translation System for Adversarial Attack and Data Augmentation Languages Experimented: Data Overview: Source Target Tra

Steven Tan 1 Aug 18, 2022
Neural Module Network for VQA in Pytorch

Neural Module Network (NMN) for VQA in Pytorch Note: This is NOT an official repository for Neural Module Networks. NMN is a network that is assembled

Harsh Trivedi 111 Nov 24, 2022
Extending JAX with custom C++ and CUDA code

Extending JAX with custom C++ and CUDA code This repository is meant as a tutorial demonstrating the infrastructure required to provide custom ops in

Dan Foreman-Mackey 237 Dec 23, 2022
Kernel Point Convolutions

Created by Hugues THOMAS Introduction Update 27/04/2020: New PyTorch implementation available. With SemanticKitti, and Windows supported. This reposit

Hugues THOMAS 584 Jan 07, 2023
DFM: A Performance Baseline for Deep Feature Matching

DFM: A Performance Baseline for Deep Feature Matching Python (Pytorch) and Matlab (MatConvNet) implementations of our paper DFM: A Performance Baselin

143 Jan 02, 2023