When BERT Plays the Lottery, All Tickets Are Winning

Overview

When BERT Plays the Lottery, All Tickets Are Winning

Large Transformer-based models were shown to be reducible to a smaller number of self-attention heads and layers. We consider this phenomenon from the perspective of the lottery ticket hypothesis, using both structured and magnitude pruning. For fine-tuned BERT, we show that (a) it is possible to find subnetworks achieving performance that is comparable with that of the full model, and (b) similarly-sized subnetworks sampled from the rest of the model perform worse. Strikingly, with structured pruning even the worst possible subnetworks remain highly trainable, indicating that most pre-trained BERT weights are potentially useful. We also study the "good" subnetworks to see if their success can be attributed to superior linguistic knowledge, but find them unstable, and not explained by meaningful self-attention patterns.

Environment

Install the requirements in your python 3.7.7 virtual environment.

pip install -r requirements.txt

These experiments were done on multi-gpu environment, were some experiments, benchmarks were run parallel. So some changes to the bash scripts to make it work for your environment.

Dataset

  1. Download the GLUE dataset using data/download_glue.py and data/download_mnli_data.py. Follow the instructions in data/download_glue.py docstring for MRPC.
  2. All data for the tasks should be organized in data/glue/task_name/ structure.
  3. Extract the attention pattern classification labelled data.
    cd data
    tar -xvf head_classification_data.tar.gz

Training, Masking, and Evaluation

Switch cwd to src (cd src) as many paths are relative from that directory.

  1. Fine-tune the BERT on GLUE tasks
./train.sh
  1. Obtain the masks
./find_masks.sh
  1. Train models with the masks applied in good, random and bad settings.
./train_with_masks.sh
  1. Evaluate the trained models
./evaluate.sh

Note: These experiments were run through course of time and now stiched together into single scripts. So it might be better to run the training and evaluation commands in them one by one.

  1. Train the CNN classifier on attention patterns normed and raw.
python classify_attention_patterns.py
python classify_normed_patterns.py

These only train the classifier.

Evaluation Analysis and Final Results

These are primarily done in jupyter notebooks in experiment_analysis directory. There are many experimental notebooks there. Here are the important ones used to generate results included in the paper.

  1. Importance pruning Heatmaps. Ignore the final "train_subset" and "hans" settings.
  2. Magnitude pruning Heatmap
  3. Overlap of surviving components
  4. Generate the random baseline
  5. Attention Classification Patterns
  6. Evaluation Result Comparisons and table
  7. Statistics on mask correlation across seeds
Owner
Sai
Machine Learning Researcher - NLP
Sai
A Lightweight Experiment & Resource Monitoring Tool 📺

Lightweight Experiment & Resource Monitoring 📺 "Did I already run this experiment before? How many resources are currently available on my cluster?"

170 Dec 28, 2022
Torch code for our CVPR 2018 paper "Residual Dense Network for Image Super-Resolution" (Spotlight)

Residual Dense Network for Image Super-Resolution This repository is for RDN introduced in the following paper Yulun Zhang, Yapeng Tian, Yu Kong, Bine

Yulun Zhang 494 Dec 30, 2022
DSTC10 Track 2 - Knowledge-grounded Task-oriented Dialogue Modeling on Spoken Conversations

DSTC10 Track 2 - Knowledge-grounded Task-oriented Dialogue Modeling on Spoken Conversations This repository contains the data, scripts and baseline co

Alexa 51 Dec 17, 2022
HODEmu, is both an executable and a python library that is based on Ragagnin 2021 in prep.

HODEmu HODEmu, is both an executable and a python library that is based on Ragagnin 2021 in prep. and emulates satellite abundance as a function of co

Antonio Ragagnin 1 Oct 13, 2021
MASA-SR: Matching Acceleration and Spatial Adaptation for Reference-Based Image Super-Resolution (CVPR2021)

MASA-SR Official PyTorch implementation of our CVPR2021 paper MASA-SR: Matching Acceleration and Spatial Adaptation for Reference-Based Image Super-Re

DV Lab 126 Dec 20, 2022
Matthew Colbrook 1 Apr 08, 2022
Multi-atlas segmentation (MAS) is a promising framework for medical image segmentation

Multi-atlas segmentation (MAS) is a promising framework for medical image segmentation. Generally, MAS methods register multiple atlases, i.e., medical images with corresponding labels, to a target i

NanYoMy 13 Oct 09, 2022
Measure WWjj polarization fraction

WlWl Polarization Measure WWjj polarization fraction Paper: arXiv:2109.09924 Notice: This code can only be used for the inference process, if you want

4 Apr 10, 2022
[NeurIPS 2021] Better Safe Than Sorry: Preventing Delusive Adversaries with Adversarial Training

Better Safe Than Sorry: Preventing Delusive Adversaries with Adversarial Training Code for NeurIPS 2021 paper "Better Safe Than Sorry: Preventing Delu

Lue Tao 29 Sep 20, 2022
Telegram chatbot created with deep learning model (LSTM) and telebot library.

Telegram chatbot Telegram chatbot created with deep learning model (LSTM) and telebot library. Description This program will allow you to create very

1 Jan 04, 2022
Patches desktop steam to look like the new steamdeck ui.

steam_deck_ui_patch The Deck UI patch will patch the regular desktop steam to look like the brand new SteamDeck UI. This patch tool currently works on

The_IT_Dude 3 Aug 29, 2022
Self-supervised Multi-modal Hybrid Fusion Network for Brain Tumor Segmentation

JBHI-Pytorch This repository contains a reference implementation of the algorithms described in our paper "Self-supervised Multi-modal Hybrid Fusion N

FeiyiFANG 5 Dec 13, 2021
An implementation of paper `Real-time Convolutional Neural Networks for Emotion and Gender Classification` with PaddlePaddle.

简介 通过PaddlePaddle框架复现了论文 Real-time Convolutional Neural Networks for Emotion and Gender Classification 中提出的两个模型,分别是SimpleCNN和MiniXception。利用 imdb_crop

8 Mar 11, 2022
Corruption Invariant Learning for Re-identification

Corruption Invariant Learning for Re-identification The official repository for Benchmarks for Corruption Invariant Person Re-identification (NeurIPS

Minghui Chen 73 Dec 08, 2022
PICARD - Parsing Incrementally for Constrained Auto-Regressive Decoding from Language Models

This is the official implementation of the following paper: Torsten Scholak, Nathan Schucher, Dzmitry Bahdanau. PICARD - Parsing Incrementally for Con

ElementAI 217 Jan 01, 2023
TANL: Structured Prediction as Translation between Augmented Natural Languages

TANL: Structured Prediction as Translation between Augmented Natural Languages Code for the paper "Structured Prediction as Translation between Augmen

98 Dec 15, 2022
Efficient electromagnetic solver based on rigorous coupled-wave analysis for 3D and 2D multi-layered structures with in-plane periodicity

Efficient electromagnetic solver based on rigorous coupled-wave analysis for 3D and 2D multi-layered structures with in-plane periodicity, such as gratings, photonic-crystal slabs, metasurfaces, surf

Alex Song 17 Dec 19, 2022
Management Dashboard for Torchserve

Torchserve Dashboard Torchserve Dashboard using Streamlit Related blog post Usage Additional Requirement: torchserve (recommended:v0.5.2) Simply run:

Ceyda Cinarel 103 Dec 10, 2022
A Rao-Blackwellized Particle Filter for 6D Object Pose Tracking

PoseRBPF: A Rao-Blackwellized Particle Filter for 6D Object Pose Tracking PoseRBPF Paper Self-supervision Paper Pose Estimation Video Robot Manipulati

NVIDIA Research Projects 107 Dec 25, 2022
Image-retrieval-baseline - MUGE Multimodal Retrieval Baseline

MUGE Multimodal Retrieval Baseline This repo is implemented based on the open_cl

47 Dec 16, 2022