An implementation of DeepMind's Relational Recurrent Neural Networks in PyTorch.

Overview

relational-rnn-pytorch

An implementation of DeepMind's Relational Recurrent Neural Networks (Santoro et al. 2018) in PyTorch.

Relational Memory Core (RMC) module is originally from official Sonnet implementation. However, currently they do not provide a full language modeling benchmark code.

This repo is a port of RMC with additional comments. It features a full-fledged word language modeling benchmark vs. traditional LSTM.

It supports any arbitrary word token-based text dataset, including WikiText-2 & WikiText-103.

Both RMC & LSTM models support adaptive softmax for much lower memory usage of large vocabulary dataset. RMC supports PyTorch's DataParallel, so you can easily experiment with a multi-GPU setup.

benchmark codes are hard-forked from official PyTorch word-language-model example

It also features an N-th farthest synthetic task from the paper (see below).

Requirements

PyTorch 0.4.1 or later (Tested on 1.0.0) & Python 3.6

Examples

python train_rmc.py --cuda for full training & test run of RMC with GPU.

python train_rmc.py --cuda --adaptivesoftmax --cutoffs 1000 5000 20000 if using large vocabulary dataset (like WikiText-103) to fit all the tensors in the VRAM.

python generate_rmc.py --cuda for generating sentences from the trained model.

python train_rnn.py --cuda for full training & test run of traditional RNN with GPU.

All default hyperparameters of RMC & LSTM are results from a two-week experiment using WikiText-2.

Data Preparation

Tested with WikiText-2 and WikiText-103. WikiText-2 is bundled.

Create a subfolder inside ./data and place word-level train.txt, valid.txt, and test.txt inside the subfolder.

Specify --data=(subfolder name) and you are good to go.

The code performs tokenization at the first training run, and the corpus is saved as pickle. The code will load the pickle file after the first run.

WikiText-2 Benchmark Results

Both RMC & LSTM have ~11M parameters. Please refer to the training code for details on hyperparameters.

Models Valid Perplexity Test Perplexity Forward pass ms/batch (TITAN Xp) Forward pass ms/batch (TITAN V)
LSTM (CuDNN) 111.31 105.56 26~27 40~41
LSTM (For Loop) Same as CuDNN Same as CuDNN 30~31 60~61
RMC 112.77 107.21 110~130 220~230

RMC can reach a comparable performance to LSTM (with heavy hyperparameter search), but it turns out that the RMC is very slow. The multi-head self-attention at every time step may be the culprit here. Using LSTMCell with for loop (which is more "fair" benchmark for RMC) slows down the forward pass, but it's still much faster.

Please also note that the hyperparameter for RMC is a worst-case scenario in terms of speed, because it used a single memory slot (as described in the paper) and did not benefit from a row-wise weight sharing from multi-slot memory.

Interesting to note here is that the speed is slower in TITAN V than TITAN Xp. The reason might be that the models are relatively small and the model calls small linear operations frequently.

Maybe TITAN Xp (~1,900Mhz unlocked CUDA clock speed vs. TITAN V's 1,335Mhz limit) benefits from these kind of workload. Or maybe TITAN V's CUDA kernel launch latency is higher for the ops in the model.

I'm not an expert in details of CUDA. Please share your results!

RMC Hyperparameter Search Results

Attention parameters tend to overfit the WikiText-2. reducing the hyperparmeters for attention (key_size) can combat the overfitting.

Applying dropout at the output logit before the softmax (like the LSTM one) helped preventing the overfitting.

embed & head size # heads attention MLP layers key size dropout at output memory slots test ppl
128 4 3 128 No 1 128.81
128 4 3 128 No 1 128.81
128 8 3 128 No 1 141.84
128 4 3 32 No 1 123.26
128 4 3 32 Yes 1 112.4
128 4 3 64 No 1 124.44
128 4 3 64 Yes 1 110.16
128 4 2 64 Yes 1 111.67
64 4 3 64 Yes 1 133.68
64 4 3 32 Yes 1 135.93
64 4 3 64 Yes 4 137.93
192 4 3 64 Yes 1 107.21
192 4 3 64 Yes 4 114.85
256 4 3 256 No 1 194.73
256 4 3 64 Yes 1 126.39

About WikiText-103

The original RMC paper presents WikiText-103 results with a larger model & batch size (6 Tesla P100, each with 64 batch size, so a total of 384. Ouch).

Using a full softmax easily blows up the VRAM. Using --adaptivesoftmax is highly recommended. If using --adaptivesoftmax, --cutoffs should be properly provided. Please refer to the original API description

I don't have such hardware and my resource is too limited to do the experiments. Benchmark result, or any other contributions are very welcome!

Nth Farthest Task

The objective of the task is: Given k randomly labelled (from 1 to k) D-dimensional vectors, identify which is the Nth farthest vector from vector M. (The answer is an integer from 1 to k.)

The specific task in the paper is: given 8 labelled 16-dimensional vectors, which is the Nth farthest vector from vector M? The vectors are labelled randomly so the model has to recognise that the Mth vector is the vector labelled as M as opposed to the vector in the Mth position in the input.

The input to the model comprises 8 40-dimensional vectors for each example. Each of these 40-dimensional vectors is structured like this:

[(vector 1) (label: which vector is it, from 1 to 8, one-hot encoded) (N, one-hot encoded) (M, one-hot encoded)] 

Example

python train_nth_farthest.py --cuda for training and testing on the Nth Farthest Task with GPU(s).

This uses the RelationalMemory class in relational_rnn_general.py, which is a version of relational_rnn_models.py without the language-modelling specific code.

Please refer totrain_nth_farthest.py for details on hyperparameter values. These are taken from Appendix A1 in the paper and from the Sonnet implementation when the hyperparameter values are not given in the paper.

Note: new examples are generated per epoch as in the Sonnet implementation. This seems to be consistent with the paper, which does not specify the number of examples used.

Experiment results

The model has been trained with a single TITAN Xp GPU for forever until it reaches 91% test accuracy. Below are the results with 3 independent runs:

The model does break the 25% barrier if trained long enough, but the wall clock time is roughly over 2~3x longer than those reported in the paper.

TODO

Experiment with different hyperparameters

Owner
Sang-gil Lee
Ph.D. student in ML/AI @ Seoul National University, South Korea. I do deep learning for sequence & generative models.
Sang-gil Lee
Food Drinks and groceries Images Multi Lingual (FooDI-ML) dataset.

Food Drinks and groceries Images Multi Lingual (FooDI-ML) dataset.

41 Jan 04, 2023
An SMPC companion library for Syft

SyMPC A library that extends PySyft with SMPC support SyMPC /ˈsɪmpəθi/ is a library which extends PySyft ≥0.3 with SMPC support. It allows computing o

Arturo Marquez Flores 0 Oct 13, 2021
Cross-modal Retrieval using Transformer Encoder Reasoning Networks (TERN). With use of Metric Learning and FAISS for fast similarity search on GPU

Cross-modal Retrieval using Transformer Encoder Reasoning Networks This project reimplements the idea from "Transformer Reasoning Network for Image-Te

Minh-Khoi Pham 5 Nov 05, 2022
UNAVOIDS: Unsupervised and Nonparametric Approach for Visualizing Outliers and Invariant Detection Scoring

UNAVOIDS: Unsupervised and Nonparametric Approach for Visualizing Outliers and Invariant Detection Scoring Code Summary aggregate.py: this script aggr

1 Dec 28, 2021
Create images and texts with the First Order Generative Adversarial Networks

First Order Divergence for training GANs This repository contains code accompanying the paper First Order Generative Advesarial Netoworks The majority

Zalando Research 35 Dec 11, 2021
A collection of IPython notebooks covering various topics.

ipython-notebooks This repo contains various IPython notebooks I've created to experiment with libraries and work through exercises, and explore subje

John Wittenauer 2.6k Jan 01, 2023
Neural-Pull: Learning Signed Distance Functions from Point Clouds by Learning to Pull Space onto Surfaces(ICML 2021)

Neural-Pull: Learning Signed Distance Functions from Point Clouds by Learning to Pull Space onto Surfaces(ICML 2021) This repository contains the code

149 Dec 15, 2022
Paddle-Adversarial-Toolbox (PAT) is a Python library for Deep Learning Security based on PaddlePaddle.

Paddle-Adversarial-Toolbox Paddle-Adversarial-Toolbox (PAT) is a Python library for Deep Learning Security based on PaddlePaddle. Model Zoo Common FGS

AgentMaker 17 Nov 08, 2022
Github for the conference paper GLOD-Gaussian Likelihood OOD detector

FOOD - Fast OOD Detector Pytorch implamentation of the confernce peper FOOD arxiv link. Abstract Deep neural networks (DNNs) perform well at classifyi

17 Jun 19, 2022
A fast, dataset-agnostic, deep visual search engine for digital art history

imgs.ai imgs.ai is a fast, dataset-agnostic, deep visual search engine for digital art history based on neural network embeddings. It utilizes modern

Fabian Offert 5 Dec 14, 2022
Self-Learned Video Rain Streak Removal: When Cyclic Consistency Meets Temporal Correspondence

In this paper, we address the problem of rain streaks removal in video by developing a self-learned rain streak removal method, which does not require any clean groundtruth images in the training pro

Yang Wenhan 44 Dec 06, 2022
StackRec: Efficient Training of Very Deep Sequential Recommender Models by Iterative Stacking

StackRec: Efficient Training of Very Deep Sequential Recommender Models by Iterative Stacking Datasets You can download datasets that have been pre-pr

25 May 29, 2022
Python-experiments - A Repository which contains python scripts to automate things and make your life easier with python

Python Experiments A Repository which contains python scripts to automate things

Vivek Kumar Singh 11 Sep 25, 2022
Knowledge Distillation Toolbox for Semantic Segmentation

SegDistill: Toolbox for Knowledge Distillation on Semantic Segmentation Networks This repo contains the supported code and configuration files for Seg

9 Dec 12, 2022
HugsVision is a easy to use huggingface wrapper for state-of-the-art computer vision

HugsVision is an open-source and easy to use all-in-one huggingface wrapper for computer vision. The goal is to create a fast, flexible and user-frien

Labrak Yanis 166 Nov 27, 2022
Transformers based fully on MLPs

Awesome MLP-based Transformers papers An up-to-date list of Transformers based fully on MLPs without attention! Why this repo? After transformers and

Fawaz Sammani 35 Dec 30, 2022
Degree-Quant: Quantization-Aware Training for Graph Neural Networks.

Degree-Quant This repo provides a clean re-implementation of the code associated with the paper Degree-Quant: Quantization-Aware Training for Graph Ne

35 Oct 07, 2022
Cascaded Pyramid Network (CPN) based on Keras (Tensorflow backend)

ML2 Takehome Project Reimplementing the paper: Cascaded Pyramid Network for Multi-Person Pose Estimation Dataset The model uses the COCO dataset which

Vo Van Tu 1 Nov 22, 2021
Unofficial TensorFlow implementation of Protein Interface Prediction using Graph Convolutional Networks.

[TensorFlow] Protein Interface Prediction using Graph Convolutional Networks Unofficial TensorFlow implementation of Protein Interface Prediction usin

YeongHyeon Park 9 Oct 25, 2022