Shortformer

This repository contains the code and the final checkpoint of the Shortformer model. This file explains how to run our experiments on the WikiText-103 dataset. Read the full paper here.

The Shortformer is a combination of two methods:

Staged Training: We first train the model on short input subsequences and then train it on longer ones. This improves both train speed and evaluation perplexity.
Position-Infused Attention + Caching: We cache previously computed subsequence representations and attend to them using Position-Infused Attention. Position-Infused Attention modifies the model so that position embeddings are not added to the word embeddings at the bottom of the network, but instead, they are added to the keys and queries in the attention sublayer (but not to the values). We show that PIA + caching vastly speeds up generation and also improves perplexity.

Staged training requires no modification to the original code. To see how we implemented the Position-Infused Attention and caching, click here. Implementing PIA and caching is very easy, and we've provided detailed comments in the code to explain what how we did it.

If you use this code or results from our paper, please cite:

@misc{press2020shortformer,
      title={Shortformer: Better Language Modeling using Shorter Inputs}, 
      author={Ofir Press and Noah A. Smith and Mike Lewis},
      year={2020},
      eprint={2012.15832},
}

Requirements and Installation

This repository is a fork of the Fairseq repository and so has the same requirements.

Once you've installed the dependencies, you can install this repository by running:

pip install --editable .

Preparing the data

To download and preprocess the data, run:

cd examples/language_model/
bash prepare-wikitext-103.sh
cd ../..


TEXT=examples/language_model/wikitext-103
python preprocess.py \
    --only-source \
    --trainpref $TEXT/wiki.train.tokens \
    --validpref $TEXT/wiki.valid.tokens \
    --testpref $TEXT/wiki.test.tokens \
    --destdir data-bin/wikitext-103 \
    --workers 20

Train/Inference for the different models

Shortformer

Our Shortformer model takes the baseline and adds caching, Position-Infused Attention, and Staged Training.

To train the first stage:

python train.py --task language_modeling     data-bin/wikitext-103     --save-dir checkpoints128e100/     --arch transformer_lm_wiki103     --max-update 140100 --max-lr 1.0 --t-mult 2 --lr-period-updates 270000 --lr-scheduler cosine --lr-shrink 0.75     --warmup-updates 16000 --warmup-init-lr 1e-07 --min-lr 1e-09 --optimizer nag --lr 0.0001 --clip-norm 0.1     --criterion adaptive_loss --max-tokens 9216 --update-freq 1 --seed 1 --fp16     --sample-break-mode none --skip-invalid-size-inputs-valid-test --ddp-backend=no_c10d --tokens-per-sample 128 --max-tokens-valid 128 --tokens-from-prev 128 --curriculum 1000 --required-batch-size-multiple 1 --save-interval 100

If your GPUs don't have enough memory to execute that command, you can set --update-freq to 2 and --max-tokens to 4608, or set --update-freq to 3 and --max-tokens to 3072 for running the model with even lower memory constraints. This chunks the batch into 2 or 3 different parts and computes each part seperately (instead of in parallel), so it uses less memory but runs slower.

After that, to train the model with the second (and final) stage:

python train.py --task language_modeling     data-bin/wikitext-103     --save-dir shortformer/ --restore-file checkpoints128e100/checkpoint100.pt     --arch transformer_lm_wiki103     --max-update 286000 --max-lr 1.0 --t-mult 2 --lr-period-updates 270000 --lr-scheduler cosine --lr-shrink 0.75     --warmup-updates 16000 --warmup-init-lr 1e-07 --min-lr 1e-09 --optimizer nag --lr 0.0001 --clip-norm 0.1     --criterion adaptive_loss --max-tokens 9216 --update-freq 1 --seed 1 --fp16     --sample-break-mode none --skip-invalid-size-inputs-valid-test --ddp-backend=no_c10d --tokens-per-sample 512 --max-tokens-valid 512 --tokens-from-prev 512 --curriculum 1000 --required-batch-size-multiple 1 --no-epoch-checkpoints

Again, you can use the update-freq/max-tokens method from above if you run out of memory.

Saved Checkpoint

If you'd like to download the Shortformer instead of training it, it is available here. Rename that file to checkpoint_best.pt if you'd like to follow the directions below.

Inference

For nonoverlapping evaluation of the validation set, run:

fairseq-eval-lm data-bin/wikitext-103     --path shortformer/checkpoint_best.pt  --sample-break-mode none --gen-subset valid   --max-sentences 1

For token-by-token generation of the validation set, run:

fairseq-eval-lm data-bin/wikitext-103     --path shortformer/checkpoint_best.pt  --sample-break-mode none --gen-subset valid   --max-sentences 1 --sliding-inf 1 --context-window 511 --max-tokens 512

(Note that --context-window is a fairseq command and doesn't have the exact meaning that the term "context window" has in our paper.)

Shortformer (without Staged Training)

Staged training improves the perplexity of the model and makes training faster, so there's no reason not to use it, but if you would like to train the Shortformer without it, the command is

python train.py --task language_modeling     data-bin/wikitext-103     --save-dir shortformer-no-st/      --arch transformer_lm_wiki103     --max-update 286000 --max-lr 1.0 --t-mult 2 --lr-period-updates 270000 --lr-scheduler cosine --lr-shrink 0.75     --warmup-updates 16000 --warmup-init-lr 1e-07 --min-lr 1e-09 --optimizer nag --lr 0.0001 --clip-norm 0.1     --criterion adaptive_loss --max-tokens 9216 --update-freq 1 --seed 1 --fp16     --sample-break-mode none --skip-invalid-size-inputs-valid-test --ddp-backend=no_c10d --tokens-per-sample 512 --max-tokens-valid 512 --tokens-from-prev 512 --curriculum 1000 --required-batch-size-multiple 1 --no-epoch-checkpoints

For inference, use the same commands as the ones for the Shortformer (above).

Baseline with Staged Training

Our Shortformer model is fast to train and for token-by-token generation, but if speed is not an issue, we can achieve slightly better performance by just applying Staged Training to the Baevski & Auli baseline LM. This model is very slow but achieves the best perplexity.

To train the first stage, download the unmodified fairseq reporsitory and then run:

python train.py --task language_modeling     data-bin/wikitext-103     --save-dir checkpoints-st-128e50/     --arch transformer_lm_wiki103     --max-update 70050 --max-lr 1.0 --t-mult 2 --lr-period-updates 270000 --lr-scheduler cosine --lr-shrink 0.75     --warmup-updates 16000 --warmup-init-lr 1e-07 --min-lr 1e-09 --optimizer nag --lr 0.0001 --clip-norm 0.1     --criterion adaptive_loss --max-tokens 9216 --update-freq 1 --seed 1 --fp16     --sample-break-mode none --skip-invalid-size-inputs-valid-test --ddp-backend=no_c10d --tokens-per-sample 128  --required-batch-size-multiple 1 --save-interval 50

After that, to train the model with the second (and final) stage:

python train.py --task language_modeling     data-bin/wikitext-103     --save-dir st/ --restore-file checkpoints-st-128e50/checkpoint50.pt     --arch transformer_lm_wiki103     --max-update 286000 --max-lr 1.0 --t-mult 2 --lr-period-updates 270000 --lr-scheduler cosine --lr-shrink 0.75     --warmup-updates 16000 --warmup-init-lr 1e-07 --min-lr 1e-09 --optimizer nag --lr 0.0001 --clip-norm 0.1     --criterion adaptive_loss --max-tokens 3072 --update-freq 3 --seed 1 --fp16     --sample-break-mode none --skip-invalid-size-inputs-valid-test --ddp-backend=no_c10d --tokens-per-sample 3072  --no-epoch-checkpoints

Inference

For nonoverlapping evaluation of the validation set, run:

fairseq-eval-lm data-bin/wikitext-103     --path st/checkpoint_best.pt  --sample-break-mode none --gen-subset valid   --max-sentences 1

For sliding window evaluation of the validation set, with a stride of 2,560, run:

fairseq-eval-lm data-bin/wikitext-103     --path st/checkpoint_best.pt  --sample-break-mode none --gen-subset valid   --max-sentences 1 --context-window 2560

Baseline - Baevski & Auli

To train the baseline, download the unmodified fairseq repository and then run:

python train.py --task language_modeling     data-bin/wikitext-103     --save-dir baseline/  --arch transformer_lm_wiki103     --max-update 286000 --max-lr 1.0 --t-mult 2 --lr-period-updates 270000 --lr-scheduler cosine --lr-shrink 0.75     --warmup-updates 16000 --warmup-init-lr 1e-07 --min-lr 1e-09 --optimizer nag --lr 0.0001 --clip-norm 0.1     --criterion adaptive_loss --max-tokens 3072 --update-freq 3 --seed 1 --fp16     --sample-break-mode none --skip-invalid-size-inputs-valid-test --ddp-backend=no_c10d --tokens-per-sample 3072  --no-epoch-checkpoints

Inference

Use the same commands as in the 'Baseline with Staged Training' inference subsection.

Code for the Shortformer model, from the paper by Ofir Press, Noah A. Smith and Mike Lewis.

Related tags

Overview

Shortformer

Requirements and Installation

Preparing the data

Train/Inference for the different models

Shortformer

Saved Checkpoint

Inference

Shortformer (without Staged Training)

Baseline with Staged Training

Inference

Baseline - Baevski & Auli

Inference

Owner

Ofir Press

Fast and Easy Infinite Neural Networks in Python

Acute ischemic stroke dataset

Visualize Camera's Pose Using Extrinsic Parameter by Plotting Pyramid Model on 3D Space

Model serving at scale

CLASP - Contrastive Language-Aminoacid Sequence Pretraining

RepMLP: Re-parameterizing Convolutions into Fully-connected Layers for Image Recognition

This is the official PyTorch implementation of our paper: "Artistic Style Transfer with Internal-external Learning and Contrastive Learning".

Cache Requests in Deta Bases and Echo them with Deta Micros

Optimal Camera Position for a Practical Application of Gaze Estimation on Edge Devices,

Unofficial PyTorch implementation of the Adaptive Convolution architecture for image style transfer

Behavioral "black-box" testing for recommender systems

CM building dataset Timisoara

This repository contains the code used for the implementation of the paper "Probabilistic Regression with HuberDistributions"

Computationally Efficient Optimization of Plackett-Luce Ranking Models for Relevance and Fairness

Modeling Category-Selective Cortical Regions with Topographic Variational Autoencoders

[ACM MM 2021] Joint Implicit Image Function for Guided Depth Super-Resolution

Scales, Chords, and Cadences: Practical Music Theory for MIR Researchers

An official implementation of MobileStyleGAN in PyTorch

ServiceX Transformer that converts flat ROOT ntuples into columnwise data

The official implementation of "Rethink Dilated Convolution for Real-time Semantic Segmentation"