Block Sparse movement pruning

Overview

Movement Pruning: Adaptive Sparsity by Fine-Tuning

Magnitude pruning is a widely used strategy for reducing model size in pure supervised learning; however, it is less effective in the transfer learning regime that has become standard for state-of-the-art natural language processing applications. We propose the use of movement pruning, a simple, deterministic first-order weight pruning method that is more adaptive to pretrained model fine-tuning. Experiments show that when pruning large pretrained language models, movement pruning shows significant improvements in high-sparsity regimes. When combined with distillation, the approach achieves minimal accuracy loss with down to only 3% of the model parameters:

Fine-pruning+Distillation
(Teacher=BERT-base fine-tuned)
BERT base
fine-tuned
Remaining
Weights (%)
Magnitude Pruning L0 Regularization Movement Pruning Soft Movement Pruning
SQuAD - Dev
EM/F1
80.4/88.1 10%
3%
70.2/80.1
45.5/59.6
72.4/81.9
64.3/75.8
75.6/84.3
67.5/78.0
76.6/84.9
72.7/82.3
MNLI - Dev
acc/MM acc
84.5/84.9 10%
3%
78.3/79.3
69.4/70.6
78.7/79.7
76.0/76.2
80.1/80.4
76.5/77.4
81.2/81.8
79.5/80.1
QQP - Dev
acc/F1
91.4/88.4 10%
3%
79.8/65.0
72.4/57.8
88.1/82.8
87.0/81.9
89.7/86.2
86.1/81.5
90.2/86.8
89.1/85.5

This page contains information on how to fine-prune pre-trained models such as BERT to obtain extremely sparse models with movement pruning. In contrast to magnitude pruning which selects weights that are far from 0, movement pruning retains weights that are moving away from 0.

For more information, we invite you to check out our paper. You can also have a look at this fun Explain Like I'm Five introductory slide deck.

Extreme sparsity and efficient storage

One promise of extreme pruning is to obtain extremely small models that can be easily sent (and stored) on edge devices. By setting weights to 0., we reduce the amount of information we need to store, and thus decreasing the memory size. We are able to obtain extremely sparse fine-pruned models with movement pruning: ~95% of the dense performance with ~5% of total remaining weights in the BERT encoder.

In this notebook, we showcase how we can leverage standard tools that exist out-of-the-box to efficiently store an extremely sparse question answering model (only 6% of total remaining weights in the encoder). We are able to reduce the memory size of the encoder from the 340MB (the orignal dense BERT) to 11MB, without any additional training of the model (every operation is performed post fine-pruning). It is sufficiently small to store it on a 91' floppy disk 📎 !

While movement pruning does not directly optimize for memory footprint (but rather the number of non-null weights), we hypothetize that further memory compression ratios can be achieved with specific quantization aware trainings (see for instance Q8BERT, And the Bit Goes Down or Quant-Noise).

Fine-pruned models

As examples, we release two English PruneBERT checkpoints (models fine-pruned from a pre-trained BERT checkpoint), one on SQuAD and the other on MNLI.

  • prunebert-base-uncased-6-finepruned-w-distil-squad
    Pre-trained BERT-base-uncased fine-pruned with soft movement pruning on SQuAD v1.1. We use an additional distillation signal from BERT-base-uncased finetuned on SQuAD. The encoder counts 6% of total non-null weights and reaches 83.8 F1 score. The model can be accessed with: pruned_bert = BertForQuestionAnswering.from_pretrained("huggingface/prunebert-base-uncased-6-finepruned-w-distil-squad")
  • prunebert-base-uncased-6-finepruned-w-distil-mnli
    Pre-trained BERT-base-uncased fine-pruned with soft movement pruning on MNLI. We use an additional distillation signal from BERT-base-uncased finetuned on MNLI. The encoder counts 6% of total non-null weights and reaches 80.7 (matched) accuracy. The model can be accessed with: pruned_bert = BertForSequenceClassification.from_pretrained("huggingface/prunebert-base-uncased-6-finepruned-w-distil-mnli")

How to fine-prune?

Setup

The code relies on the 🤗 Transformers library. In addition to the dependencies listed in the examples folder, you should install a few additional dependencies listed in the requirements.txt file: pip install -r requirements.txt.

Note that we built our experiments on top of a stabilized version of the library (commit https://github.com/huggingface/transformers/commit/352d5472b0c1dec0f420d606d16747d851b4bda8): we do not guarantee that everything is still compatible with the latest version of the master branch.

Fine-pruning with movement pruning

Below, we detail how to reproduce the results reported in the paper. We use SQuAD as a running example. Commands (and scripts) can be easily adapted for other tasks.

The following command fine-prunes a pre-trained BERT-base on SQuAD using movement pruning towards 15% of remaining weights (85% sparsity). Note that we freeze all the embeddings modules (from their pre-trained value) and only prune the Fully Connected layers in the encoder (12 layers of Transformer Block).

SERIALIZATION_DIR=<OUTPUT_DIR>
SQUAD_DATA=squad_data

mkdir $SQUAD_DATA
cd $SQUAD_DATA
wget -q https://rajpurkar.github.io/SQuAD-explorer/dataset/dev-v1.1.json
wget -q https://rajpurkar.github.io/SQuAD-explorer/dataset/train-v1.1.json
cd ..


python examples/movement-pruning/masked_run_squad.py \
    --output_dir $SERIALIZATION_DIR \
    --data_dir $SQUAD_DATA \
    --train_file train-v1.1.json \
    --predict_file dev-v1.1.json \
    --do_train --do_eval --do_lower_case \
    --model_type masked_bert \
    --model_name_or_path bert-base-uncased \
    --per_gpu_train_batch_size 16 \
    --warmup_steps 5400 \
    --num_train_epochs 10 \
    --learning_rate 3e-5 --mask_scores_learning_rate 1e-2 \
    --initial_threshold 1 --final_threshold 0.15 \
    --initial_warmup 1 --final_warmup 2 \
    --pruning_method topK --mask_init constant --mask_scale 0.

Fine-pruning with other methods

We can also explore other fine-pruning methods by changing the pruning_method parameter:

Soft movement pruning

python examples/movement-pruning/masked_run_squad.py \
    --output_dir $SERIALIZATION_DIR \
    --data_dir $SQUAD_DATA \
    --train_file train-v1.1.json \
    --predict_file dev-v1.1.json \
    --do_train --do_eval --do_lower_case \
    --model_type masked_bert \
    --model_name_or_path bert-base-uncased \
    --per_gpu_train_batch_size 16 \
    --warmup_steps 5400 \
    --num_train_epochs 10 \
    --learning_rate 3e-5 --mask_scores_learning_rate 1e-2 \
    --initial_threshold 0 --final_threshold 0.1 \
    --initial_warmup 1 --final_warmup 2 \
    --pruning_method sigmoied_threshold --mask_init constant --mask_scale 0. \
    --regularization l1 --final_lambda 400.

L0 regularization

python examples/movement-pruning/masked_run_squad.py \
    --output_dir $SERIALIZATION_DIR \
    --data_dir $SQUAD_DATA \
    --train_file train-v1.1.json \
    --predict_file dev-v1.1.json \
    --do_train --do_eval --do_lower_case \
    --model_type masked_bert \
    --model_name_or_path bert-base-uncased \
    --per_gpu_train_batch_size 16 \
    --warmup_steps 5400 \
    --num_train_epochs 10 \
    --learning_rate 3e-5 --mask_scores_learning_rate 1e-1 \
    --initial_threshold 1. --final_threshold 1. \
    --initial_warmup 1 --final_warmup 1 \
    --pruning_method l0 --mask_init constant --mask_scale 2.197 \
    --regularization l0 --final_lambda 125.

Iterative Magnitude Pruning

python examples/movement-pruning/masked_run_squad.py \
    --output_dir ./dbg \
    --data_dir examples/distillation/data/squad_data \
    --train_file train-v1.1.json \
    --predict_file dev-v1.1.json \
    --do_train --do_eval --do_lower_case \
    --model_type masked_bert \
    --model_name_or_path bert-base-uncased \
    --per_gpu_train_batch_size 16 \
    --warmup_steps 5400 \
    --num_train_epochs 10 \
    --learning_rate 3e-5 \
    --initial_threshold 1 --final_threshold 0.15 \
    --initial_warmup 1 --final_warmup 2 \
    --pruning_method magnitude

After fine-pruning

Counting parameters

Regularization based pruning methods (soft movement pruning and L0 regularization) rely on the penalty to induce sparsity. The multiplicative coefficient controls the sparsity level. To obtain the effective sparsity level in the encoder, we simply count the number of activated (non-null) weights:

python examples/movement-pruning/counts_parameters.py \
    --pruning_method sigmoied_threshold \
    --threshold 0.1 \
    --serialization_dir $SERIALIZATION_DIR

Pruning once for all

Once the model has been fine-pruned, the pruned weights can be set to 0. once for all (reducing the amount of information to store). In our running experiments, we can convert a MaskedBertForQuestionAnswering (a BERT model augmented to enable on-the-fly pruning capabilities) to a standard BertForQuestionAnswering:

python examples/movement-pruning/bertarize.py \
    --pruning_method sigmoied_threshold \
    --threshold 0.1 \
    --model_name_or_path $SERIALIZATION_DIR

Hyper-parameters

For reproducibility purposes, we share the detailed results presented in the paper. These tables exhaustively describe the individual hyper-parameters used for each data point.

Inference speed

Early experiments show that even though models fine-pruned with (soft) movement pruning are extremely sparse, they do not benefit from significant improvement in terms of inference speed when using the standard PyTorch inference. We are currently benchmarking and exploring inference setups specifically for sparse architectures. In particular, hardware manufacturers are announcing devices that will speedup inference for sparse networks considerably.

Citation

If you find this resource useful, please consider citing the following paper:

@article{sanh2020movement,
    title={Movement Pruning: Adaptive Sparsity by Fine-Tuning},
    author={Victor Sanh and Thomas Wolf and Alexander M. Rush},
    year={2020},
    eprint={2005.07683},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}
Owner
Hugging Face
Solving NLP, one commit at a time!
Hugging Face
Two-Stream Adaptive Graph Convolutional Networks for Skeleton-Based Action Recognition in CVPR19

2s-AGCN Two-Stream Adaptive Graph Convolutional Networks for Skeleton-Based Action Recognition in CVPR19 Note PyTorch version should be 0.3! For PyTor

LShi 547 Dec 26, 2022
Code for Ditto: Building Digital Twins of Articulated Objects from Interaction

Ditto: Building Digital Twins of Articulated Objects from Interaction Zhenyu Jiang, Cheng-Chun Hsu, Yuke Zhu CVPR 2022, Oral Project | arxiv News 2022

UT Robot Perception and Learning Lab 78 Dec 22, 2022
The implementation of the CVPR2021 paper "Structure-Aware Face Clustering on a Large-Scale Graph with 10^7 Nodes"

STAR-FC This code is the implementation for the CVPR 2021 paper "Structure-Aware Face Clustering on a Large-Scale Graph with 10^7 Nodes" 🌟 🌟 . 🎓 Re

Shuai Shen 87 Dec 28, 2022
Parameterising Simulated Annealing for the Travelling Salesman Problem

Parameterising Simulated Annealing for the Travelling Salesman Problem

Gary Sun 55 Jun 15, 2022
一个目标检测的通用框架(不需要cuda编译),支持Yolo全系列(v2~v5)、EfficientDet、RetinaNet、Cascade-RCNN等SOTA网络。

一个目标检测的通用框架(不需要cuda编译),支持Yolo全系列(v2~v5)、EfficientDet、RetinaNet、Cascade-RCNN等SOTA网络。

Haoyu Xu 203 Jan 03, 2023
Implementation of paper "Graph Condensation for Graph Neural Networks"

GCond A PyTorch implementation of paper "Graph Condensation for Graph Neural Networks" Code will be released soon. Stay tuned :) Abstract We propose a

Wei Jin 66 Dec 04, 2022
Rethinking Transformer-based Set Prediction for Object Detection

Rethinking Transformer-based Set Prediction for Object Detection Here are the code for the ICCV paper. The code is adapted from Detectron2 and AdelaiD

Zhiqing Sun 62 Dec 03, 2022
Share a benchmark that can easily apply reinforcement learning in Job-shop-scheduling

Gymjsp Gymjsp is an open source Python library, which uses the OpenAI Gym interface for easily instantiating and interacting with RL environments, and

134 Dec 08, 2022
Ranger deep learning optimizer rewrite to use newest components

Ranger21 - integrating the latest deep learning components into a single optimizer Ranger deep learning optimizer rewrite to use newest components Ran

Less Wright 266 Dec 28, 2022
A framework for the elicitation, specification, formalization and understanding of requirements.

A framework for the elicitation, specification, formalization and understanding of requirements.

NASA - Software V&V 161 Jan 03, 2023
load .txt to train YOLOX, same as Yolo others

YOLOX train your data you need generate data.txt like follow format (per line- one image). prepare one data.txt like this: img_path1 x1,y1,x2,y2,clas

LiMingf 18 Aug 18, 2022
CT-Net: Channel Tensorization Network for Video Classification

[ICLR2021] CT-Net: Channel Tensorization Network for Video Classification @inproceedings{ li2021ctnet, title={{\{}CT{\}}-Net: Channel Tensorization Ne

33 Nov 15, 2022
Code for Estimating Multi-cause Treatment Effects via Single-cause Perturbation (NeurIPS 2021)

Estimating Multi-cause Treatment Effects via Single-cause Perturbation (NeurIPS 2021) Single-cause Perturbation (SCP) is a framework to estimate the m

Zhaozhi Qian 9 Sep 28, 2022
Automatically creates genre collections for your Plex media

Plex Auto Genres Plex Auto Genres is a simple script that will add genre collection tags to your media making it much easier to search for genre speci

Shane Israel 63 Dec 31, 2022
IOT: Instance-wise Layer Reordering for Transformer Structures

Introduction This repository contains the code for Instance-wise Ordered Transformer (IOT), which is introduced in the ICLR2021 paper IOT: Instance-wi

IOT 19 Nov 15, 2022
This is the official implementation of VaxNeRF (Voxel-Accelearated NeRF).

VaxNeRF Paper | Google Colab This is the official implementation of VaxNeRF (Voxel-Accelearated NeRF). This codebase is implemented using JAX, buildin

naruya 132 Nov 21, 2022
Simple ONNX operation generator. Simple Operation Generator for ONNX.

sog4onnx Simple ONNX operation generator. Simple Operation Generator for ONNX. https://github.com/PINTO0309/simple-onnx-processing-tools Key concept V

Katsuya Hyodo 6 May 15, 2022
OMNIVORE is a single vision model for many different visual modalities

Omnivore: A Single Model for Many Visual Modalities [paper][website] OMNIVORE is a single vision model for many different visual modalities. It learns

Meta Research 451 Dec 27, 2022
System Design course at HSE (2021)

System Design course at HSE (2021) Wiki-страница курса Структура репозитория: slides - директория с презентациями с занятий tasks - материалы для выпо

22 Dec 25, 2022
Autoencoders pretraining using clustering

Autoencoders pretraining using clustering

IITiS PAN 2 Dec 16, 2021