Optimizing DR with hard negatives and achieving SOTA first-stage retrieval performance on TREC DL Track (SIGIR 2021 Full Paper).

Overview

Optimizing Dense Retrieval Model Training with Hard Negatives

Jingtao Zhan, Jiaxin Mao, Yiqun Liu, Jiafeng Guo, Min Zhang, Shaoping Ma

This repo provides code, retrieval results, and trained models for our SIGIR Full paper Optimizing Dense Retrieval Model Training with Hard Negatives. The previous version is Learning To Retrieve: How to Train a Dense Retrieval Model Effectively and Efficiently.

We achieve very impressive retrieval results on both passage and document retrieval bechmarks. The proposed two algorithms (STAR and ADORE) are very efficient. IMHO, they are well worth trying and most likely improve your retriever's performance by a large margin.

The following figure shows the pros and cons of different training methods. You can train an effective Dense Retrieval model in three steps. Firstly, warmup your model using random negatives or BM25 top negatives. Secondly, use our proposed STAR to train the query encoder and document encoder. Thirdly, use our proposed ADORE to train the query encoder. image

Retrieval Results and Trained Models

Passage Retrieval Dev [email protected] Dev [email protected] Test [email protected] Files
Inbatch-Neg 0.264 0.837 0.583 Model
Rand-Neg 0.301 0.853 0.612 Model
STAR 0.340 0.867 0.642 Model Train Dev TRECTest
ADORE (Inbatch-Neg) 0.316 0.860 0.658 Model
ADORE (Rand-Neg) 0.326 0.865 0.661 Model
ADORE (STAR) 0.347 0.876 0.683 Model Train Dev TRECTest Leaderboard
Doc Retrieval Dev [email protected] Dev [email protected] Test [email protected] Files
Inbatch-Neg 0.320 0.864 0.544 Model
Rand-Neg 0.330 0.859 0.572 Model
STAR 0.390 0.867 0.605 Model Train Dev TRECTest
ADORE (Inbatch-Neg) 0.362 0.884 0.580 Model
ADORE (Rand-Neg) 0.361 0.885 0.585 Model
ADORE (STAR) 0.405 0.919 0.628 Model Train Dev TRECTest Leaderboard

If you want to use our first-stage leaderboard runs, contact me and I will send you the file.

If any links fail or the files go wrong, please contact me or open a issue.

Requirements

To install requirements, run the following commands:

git clone [email protected]:jingtaozhan/DRhard.git
cd DRhard
python setup.py install

However, you need to set up a new python enverionment for data preprocessing (see below).

Data Download

To download all the needed data, run:

bash download_data.sh

Data Preprocess

You need to set up a new environment with transformers==2.8.0 to tokenize the text. This is because we find the tokenizer behaves differently among versions 2, 3 and 4. To replicate the results in our paper with our provided trained models, it is necessary to use version 2.8.0 for preprocessing. Otherwise, you may need to re-train the DR models.

Run the following codes.

python preprocess.py --data_type 0; python preprocess.py --data_type 1

Inference

With our provided trained models, you can easily replicate our reported experimental results. Note that minor variance may be observed due to environmental difference.

STAR

The following codes use the provided STAR model to compute query/passage embeddings and perform similarity search on the dev set. (You can use --faiss_gpus option to use gpus for much faster similarity search.)

python ./star/inference.py --data_type passage --max_doc_length 256 --mode dev   
python ./star/inference.py --data_type doc --max_doc_length 512 --mode dev   

Run the following code to evaluate on MSMARCO Passage dataset.

python ./msmarco_eval.py ./data/passage/preprocess/dev-qrel.tsv ./data/passage/evaluate/star/dev.rank.tsv
Eval Started
#####################
MRR @10: 0.3404237731386721
QueriesRanked: 6980
#####################

Run the following code to evaluate on MSMARCO Document dataset.

python ./msmarco_eval.py ./data/doc/preprocess/dev-qrel.tsv ./data/doc/evaluate/star/dev.rank.tsv 100
Eval Started
#####################
MRR @100: 0.3903422772218344
QueriesRanked: 5193
#####################

ADORE

ADORE computes the query embeddings. The document embeddings are pre-computed by other DR models, like STAR. The following codes use the provided ADORE(STAR) model to compute query embeddings and perform similarity search on the dev set. (You can use --faiss_gpus option to use gpus for much faster similarity search.)

python ./adore/inference.py --model_dir ./data/passage/trained_models/adore-star --output_dir ./data/passage/evaluate/adore-star --preprocess_dir ./data/passage/preprocess --mode dev --dmemmap_path ./data/passage/evaluate/star/passages.memmap
python ./adore/inference.py --model_dir ./data/doc/trained_models/adore-star --output_dir ./data/doc/evaluate/adore-star --preprocess_dir ./data/doc/preprocess --mode dev --dmemmap_path ./data/doc/evaluate/star/passages.memmap

Evaluate ADORE(STAR) model on dev passage dataset:

python ./msmarco_eval.py ./data/passage/preprocess/dev-qrel.tsv ./data/passage/evaluate/adore-star/dev.rank.tsv

You will get

Eval Started
#####################
MRR @10: 0.34660697230181425
QueriesRanked: 6980
#####################

Evaluate ADORE(STAR) model on dev document dataset:

python ./msmarco_eval.py ./data/doc/preprocess/dev-qrel.tsv ./data/doc/evaluate/adore-star/dev.rank.tsv 100

You will get

Eval Started
#####################
MRR @100: 0.4049777020859768
QueriesRanked: 5193
#####################

Convert QID/PID Back

Our data preprocessing reassigns new ids for each query and document. Therefore, you may want to convert the ids back. We provide a script for this.

The following code shows an example to convert ADORE-STAR's ranking results on the dev passage dataset.

python ./cvt_back.py --input_dir ./data/passage/evaluate/adore-star/ --preprocess_dir ./data/passage/preprocess --output_dir ./data/passage/official_runs/adore-star --mode dev --dataset passage
python ./msmarco_eval.py ./data/passage/dataset/qrels.dev.small.tsv ./data/passage/official_runs/adore-star/dev.rank.tsv

You will get

Eval Started
#####################
MRR @10: 0.34660697230181425
QueriesRanked: 6980
#####################

Train

In the following instructions, we show how to replicate our experimental results on MSMARCO Passage Retrieval task.

STAR

We use the same warmup model as ANCE, the most competitive baseline, to enable a fair comparison. Please download it and extract it at ./data/passage/warmup

Next, we use this warmup model to extract static hard negatives, which will be utilized by STAR.

python ./star/prepare_hardneg.py \
--data_type passage \
--max_query_length 32 \
--max_doc_length 256 \
--mode dev \
--topk 200

It will automatically use all available gpus to retrieve documents. If all available cuda memory is less than 26GB (the index size), you can add --not_faiss_cuda to use CPU for retrieval.

Run the following command to train the DR model with STAR. In our experiments, we only use one GPU to train.

python ./star/train.py --do_train \
    --max_query_length 24 \
    --max_doc_length 120 \
    --preprocess_dir ./data/passage/preprocess \
    --hardneg_path ./data/passage/warmup_retrieve/hard.json \
    --init_path ./data/passage/warmup \
    --output_dir ./data/passage/star_train/models \
    --logging_dir ./data/passage/star_train/log \
    --optimizer_str lamb \
    --learning_rate 1e-4 \
    --gradient_checkpointing --fp16

Although we set number of training epcohs a very large value in the script, it is likely to converge within 50k steps (1.5 days) and you can manually kill the process. Using multiple gpus should speed up a lot, which requires some changes in the codes.

ADORE

Now we show how to use ADORE to finetune the query encoder. Here we use our provided STAR checkpoint as the fixed document encoder. You can also use another document encoder.

The passage embeddings by STAR should be located at ./data/passage/evaluate/star/passages.memmap. If not, follow the STAR inference procedure as shown above.

python ./adore/train.py \
--metric_cut 200 \
--init_path ./data/passage/trained_models/star \
--pembed_path ./data/passage/evaluate/star/passages.memmap \
--model_save_dir ./data/passage/adore_train/models \
--log_dir ./data/passage/adore_train/log \
--preprocess_dir ./data/passage/preprocess \
--model_gpu_index 0 \
--faiss_gpu_index 1 2 3

The above command uses the first gpu for encoding, and the 2nd~4th gpu for dense retrieval. You can change the faiss_gpu_index values based on your available cuda memory. For example, if you have a 32GB gpu, you can set model_gpu_index and faiss_gpu_index both to 0 because the CUDA memory is large enough. But if you only have 11GB gpus, three gpus are required for faiss.

Empirically, ADORE significantly improves retrieval performance after training for only one epoch, which only costs 1 hour if using GPUs to retrieve dynamic hard negatives.

Owner
Jingtao Zhan
IR Researcher, Ph.D student at Tsinghua University.
Jingtao Zhan
Unofficial PyTorch implementation of the Adaptive Convolution architecture for image style transfer

AdaConv Unofficial PyTorch implementation of the Adaptive Convolution architecture for image style transfer from "Adaptive Convolutions for Structure-

65 Dec 22, 2022
YOLOX + ROS(1, 2) object detection package

YOLOX + ROS(1, 2) object detection package

Ar-Ray 158 Dec 21, 2022
This repository contains the code for "SBEVNet: End-to-End Deep Stereo Layout Estimation" paper by Divam Gupta, Wei Pu, Trenton Tabor, Jeff Schneider

SBEVNet: End-to-End Deep Stereo Layout Estimation This repository contains the code for "SBEVNet: End-to-End Deep Stereo Layout Estimation" paper by D

Divam Gupta 19 Dec 17, 2022
Official Repo for Ground-aware Monocular 3D Object Detection for Autonomous Driving

Visual 3D Detection Package: This repo aims to provide flexible and reproducible visual 3D detection on KITTI dataset. We expect scripts starting from

Yuxuan Liu 305 Dec 19, 2022
Hand Gesture Volume Control is AIML based project which uses image processing to control the volume of your Computer.

Hand Gesture Volume Control Modules There are basically three modules Handtracking Program Handtracking Module Volume Control Program Handtracking Pro

VITTAL 1 Jan 12, 2022
abess: Fast Best-Subset Selection in Python and R

abess: Fast Best-Subset Selection in Python and R Overview abess (Adaptive BEst Subset Selection) library aims to solve general best subset selection,

297 Dec 21, 2022
IsoGCN code for ICLR2021

IsoGCN The official implementation of IsoGCN, presented in the ICLR2021 paper Isometric Transformation Invariant and Equivariant Graph Convolutional N

horiem 39 Nov 25, 2022
A collection of Jupyter notebooks to play with NVIDIA's StyleGAN3 and OpenAI's CLIP for a text-based guided image generation.

StyleGAN3 CLIP-based guidance StyleGAN3 + CLIP StyleGAN3 + inversion + CLIP This repo is a collection of Jupyter notebooks made to easily play with St

Eugenio Herrera 176 Dec 30, 2022
Dynamic wallpaper generator.

Wiki • About • Installation About This project is a dynamic wallpaper changer. It waits untill you turn on the music, downloads album cover if it's po

3 Sep 18, 2021
Source code for CVPR 2021 paper "Riggable 3D Face Reconstruction via In-Network Optimization"

Riggable 3D Face Reconstruction via In-Network Optimization Source code for CVPR 2021 paper "Riggable 3D Face Reconstruction via In-Network Optimizati

130 Jan 02, 2023
Starter code for the ICCV 2021 paper, 'Detecting Invisible People'

Detecting Invisible People [ICCV 2021 Paper] [Website] Tarasha Khurana, Achal Dave, Deva Ramanan Introduction This repository contains code for Detect

Tarasha Khurana 28 Sep 16, 2022
DECAF: Generating Fair Synthetic Data Using Causally-Aware Generative Networks

DECAF (DEbiasing CAusal Fairness) Code Author: Trent Kyono This repository contains the code used for the "DECAF: Generating Fair Synthetic Data Using

van_der_Schaar \LAB 7 Nov 24, 2022
Source code for paper "Deep Superpixel-based Network for Blind Image Quality Assessment"

DSN-IQA Source code for paper "Deep Superpixel-based Network for Blind Image Quality Assessment" Requirements Python =3.8.0 Pytorch =1.7.1 Usage wit

7 Oct 13, 2022
Fine-tune pretrained Convolutional Neural Networks with PyTorch

Fine-tune pretrained Convolutional Neural Networks with PyTorch. Features Gives access to the most popular CNN architectures pretrained on ImageNet. A

Alex Parinov 694 Nov 23, 2022
Dyalog-apl-docset - Dyalog APL Dash Docset Generator

Dyalog APL Dash Docset Generator o alasa e kili sona kepeken tenpo lili a A Dash

Maciej Goszczycki 1 Jan 10, 2022
Pytorch reimplementation of PSM-Net: "Pyramid Stereo Matching Network"

This is a Pytorch Lightning version PSMNet which is based on JiaRenChang/PSMNet. use python main.py to start training. PSM-Net Pytorch reimplementatio

XIAOTIAN LIU 1 Nov 25, 2021
CNN Based Meta-Learning for Noisy Image Classification and Template Matching

CNN Based Meta-Learning for Noisy Image Classification and Template Matching Introduction This master thesis used a few-shot meta learning approach to

Kumar Manas 2 Dec 09, 2021
ICLR21 Tent: Fully Test-Time Adaptation by Entropy Minimization

⛺️ Tent: Fully Test-Time Adaptation by Entropy Minimization This is the official project repository for Tent: Fully-Test Time Adaptation by Entropy Mi

Dequan Wang 204 Dec 25, 2022
Framework that uses artificial intelligence applied to mathematical models to make predictions

LiconIA Framework that uses artificial intelligence applied to mathematical models to make predictions Interface Overview Table of contents [TOC] 1 Ar

4 Jun 20, 2021
A project to make Amazon Echo respond to sign language using your webcam

Making Alexa respond to Sign Language using Tensorflow.js Try the live demo Read the Blog Post on Tensorflow's Blog Coming Soon Watch the video This p

Abhishek Singh 444 Jan 03, 2023