ALBERT: A Lite BERT for Self-supervised Learning of Language Representations

Related tags

Deep Learningalbert
Overview

ALBERT

***************New March 28, 2020 ***************

Add a colab tutorial to run fine-tuning for GLUE datasets.

***************New January 7, 2020 ***************

v2 TF-Hub models should be working now with TF 1.15, as we removed the native Einsum op from the graph. See updated TF-Hub links below.

***************New December 30, 2019 ***************

Chinese models are released. We would like to thank CLUE team for providing the training data.

Version 2 of ALBERT models is released.

In this version, we apply 'no dropout', 'additional training data' and 'long training time' strategies to all models. We train ALBERT-base for 10M steps and other models for 3M steps.

The result comparison to the v1 models is as followings:

Average SQuAD1.1 SQuAD2.0 MNLI SST-2 RACE
V2
ALBERT-base 82.3 90.2/83.2 82.1/79.3 84.6 92.9 66.8
ALBERT-large 85.7 91.8/85.2 84.9/81.8 86.5 94.9 75.2
ALBERT-xlarge 87.9 92.9/86.4 87.9/84.1 87.9 95.4 80.7
ALBERT-xxlarge 90.9 94.6/89.1 89.8/86.9 90.6 96.8 86.8
V1
ALBERT-base 80.1 89.3/82.3 80.0/77.1 81.6 90.3 64.0
ALBERT-large 82.4 90.6/83.9 82.3/79.4 83.5 91.7 68.5
ALBERT-xlarge 85.5 92.5/86.1 86.1/83.1 86.4 92.4 74.8
ALBERT-xxlarge 91.0 94.8/89.3 90.2/87.4 90.8 96.9 86.5

The comparison shows that for ALBERT-base, ALBERT-large, and ALBERT-xlarge, v2 is much better than v1, indicating the importance of applying the above three strategies. On average, ALBERT-xxlarge is slightly worse than the v1, because of the following two reasons: 1) Training additional 1.5 M steps (the only difference between these two models is training for 1.5M steps and 3M steps) did not lead to significant performance improvement. 2) For v1, we did a little bit hyperparameter search among the parameters sets given by BERT, Roberta, and XLnet. For v2, we simply adopt the parameters from v1 except for RACE, where we use a learning rate of 1e-5 and 0 ALBERT DR (dropout rate for ALBERT in finetuning). The original (v1) RACE hyperparameter will cause model divergence for v2 models. Given that the downstream tasks are sensitive to the fine-tuning hyperparameters, we should be careful about so called slight improvements.

ALBERT is "A Lite" version of BERT, a popular unsupervised language representation learning algorithm. ALBERT uses parameter-reduction techniques that allow for large-scale configurations, overcome previous memory limitations, and achieve better behavior with respect to model degradation.

For a technical description of the algorithm, see our paper:

ALBERT: A Lite BERT for Self-supervised Learning of Language Representations

Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, Radu Soricut

Release Notes

  • Initial release: 10/9/2019

Results

Performance of ALBERT on GLUE benchmark results using a single-model setup on dev:

Models MNLI QNLI QQP RTE SST MRPC CoLA STS
BERT-large 86.6 92.3 91.3 70.4 93.2 88.0 60.6 90.0
XLNet-large 89.8 93.9 91.8 83.8 95.6 89.2 63.6 91.8
RoBERTa-large 90.2 94.7 92.2 86.6 96.4 90.9 68.0 92.4
ALBERT (1M) 90.4 95.2 92.0 88.1 96.8 90.2 68.7 92.7
ALBERT (1.5M) 90.8 95.3 92.2 89.2 96.9 90.9 71.4 93.0

Performance of ALBERT-xxl on SQuaD and RACE benchmarks using a single-model setup:

Models SQuAD1.1 dev SQuAD2.0 dev SQuAD2.0 test RACE test (Middle/High)
BERT-large 90.9/84.1 81.8/79.0 89.1/86.3 72.0 (76.6/70.1)
XLNet 94.5/89.0 88.8/86.1 89.1/86.3 81.8 (85.5/80.2)
RoBERTa 94.6/88.9 89.4/86.5 89.8/86.8 83.2 (86.5/81.3)
UPM - - 89.9/87.2 -
XLNet + SG-Net Verifier++ - - 90.1/87.2 -
ALBERT (1M) 94.8/89.2 89.9/87.2 - 86.0 (88.2/85.1)
ALBERT (1.5M) 94.8/89.3 90.2/87.4 90.9/88.1 86.5 (89.0/85.5)

Pre-trained Models

TF-Hub modules are available:

Example usage of the TF-Hub module in code:

tags = set()
if is_training:
  tags.add("train")
albert_module = hub.Module("https://tfhub.dev/google/albert_base/1", tags=tags,
                           trainable=True)
albert_inputs = dict(
    input_ids=input_ids,
    input_mask=input_mask,
    segment_ids=segment_ids)
albert_outputs = albert_module(
    inputs=albert_inputs,
    signature="tokens",
    as_dict=True)

# If you want to use the token-level output, use
# albert_outputs["sequence_output"] instead.
output_layer = albert_outputs["pooled_output"]

Most of the fine-tuning scripts in this repository support TF-hub modules via the --albert_hub_module_handle flag.

Pre-training Instructions

To pretrain ALBERT, use run_pretraining.py:

pip install -r albert/requirements.txt
python -m albert.run_pretraining \
    --input_file=... \
    --output_dir=... \
    --init_checkpoint=... \
    --albert_config_file=... \
    --do_train \
    --do_eval \
    --train_batch_size=4096 \
    --eval_batch_size=64 \
    --max_seq_length=512 \
    --max_predictions_per_seq=20 \
    --optimizer='lamb' \
    --learning_rate=.00176 \
    --num_train_steps=125000 \
    --num_warmup_steps=3125 \
    --save_checkpoints_steps=5000

Fine-tuning on GLUE

To fine-tune and evaluate a pretrained ALBERT on GLUE, please see the convenience script run_glue.sh.

Lower-level use cases may want to use the run_classifier.py script directly. The run_classifier.py script is used both for fine-tuning and evaluation of ALBERT on individual GLUE benchmark tasks, such as MNLI:

pip install -r albert/requirements.txt
python -m albert.run_classifier \
  --data_dir=... \
  --output_dir=... \
  --init_checkpoint=... \
  --albert_config_file=... \
  --spm_model_file=... \
  --do_train \
  --do_eval \
  --do_predict \
  --do_lower_case \
  --max_seq_length=128 \
  --optimizer=adamw \
  --task_name=MNLI \
  --warmup_step=1000 \
  --learning_rate=3e-5 \
  --train_step=10000 \
  --save_checkpoints_steps=100 \
  --train_batch_size=128

Good default flag values for each GLUE task can be found in run_glue.sh.

You can fine-tune the model starting from TF-Hub modules instead of raw checkpoints by setting e.g. --albert_hub_module_handle=https://tfhub.dev/google/albert_base/1 instead of --init_checkpoint.

You can find the spm_model_file in the tar files or under the assets folder of the tf-hub module. The name of the model file is "30k-clean.model".

After evaluation, the script should report some output like this:

***** Eval results *****
  global_step = ...
  loss = ...
  masked_lm_accuracy = ...
  masked_lm_loss = ...
  sentence_order_accuracy = ...
  sentence_order_loss = ...

Fine-tuning on SQuAD

To fine-tune and evaluate a pretrained model on SQuAD v1, use the run_squad_v1.py script:

pip install -r albert/requirements.txt
python -m albert.run_squad_v1 \
  --albert_config_file=... \
  --output_dir=... \
  --train_file=... \
  --predict_file=... \
  --train_feature_file=... \
  --predict_feature_file=... \
  --predict_feature_left_file=... \
  --init_checkpoint=... \
  --spm_model_file=... \
  --do_lower_case \
  --max_seq_length=384 \
  --doc_stride=128 \
  --max_query_length=64 \
  --do_train=true \
  --do_predict=true \
  --train_batch_size=48 \
  --predict_batch_size=8 \
  --learning_rate=5e-5 \
  --num_train_epochs=2.0 \
  --warmup_proportion=.1 \
  --save_checkpoints_steps=5000 \
  --n_best_size=20 \
  --max_answer_length=30

You can fine-tune the model starting from TF-Hub modules instead of raw checkpoints by setting e.g. --albert_hub_module_handle=https://tfhub.dev/google/albert_base/1 instead of --init_checkpoint.

For SQuAD v2, use the run_squad_v2.py script:

pip install -r albert/requirements.txt
python -m albert.run_squad_v2 \
  --albert_config_file=... \
  --output_dir=... \
  --train_file=... \
  --predict_file=... \
  --train_feature_file=... \
  --predict_feature_file=... \
  --predict_feature_left_file=... \
  --init_checkpoint=... \
  --spm_model_file=... \
  --do_lower_case \
  --max_seq_length=384 \
  --doc_stride=128 \
  --max_query_length=64 \
  --do_train \
  --do_predict \
  --train_batch_size=48 \
  --predict_batch_size=8 \
  --learning_rate=5e-5 \
  --num_train_epochs=2.0 \
  --warmup_proportion=.1 \
  --save_checkpoints_steps=5000 \
  --n_best_size=20 \
  --max_answer_length=30

You can fine-tune the model starting from TF-Hub modules instead of raw checkpoints by setting e.g. --albert_hub_module_handle=https://tfhub.dev/google/albert_base/1 instead of --init_checkpoint.

Fine-tuning on RACE

For RACE, use the run_race.py script:

pip install -r albert/requirements.txt
python -m albert.run_race \
  --albert_config_file=... \
  --output_dir=... \
  --train_file=... \
  --eval_file=... \
  --data_dir=...\
  --init_checkpoint=... \
  --spm_model_file=... \
  --max_seq_length=512 \
  --max_qa_length=128 \
  --do_train \
  --do_eval \
  --train_batch_size=32 \
  --eval_batch_size=8 \
  --learning_rate=1e-5 \
  --train_step=12000 \
  --warmup_step=1000 \
  --save_checkpoints_steps=100

You can fine-tune the model starting from TF-Hub modules instead of raw checkpoints by setting e.g. --albert_hub_module_handle=https://tfhub.dev/google/albert_base/1 instead of --init_checkpoint.

SentencePiece

Command for generating the sentence piece vocabulary:

spm_train \
--input all.txt --model_prefix=30k-clean --vocab_size=30000 --logtostderr
--pad_id=0 --unk_id=1 --eos_id=-1 --bos_id=-1
--control_symbols=[CLS],[SEP],[MASK]
--user_defined_symbols="(,),\",-,.,–,£,€"
--shuffle_input_sentence=true --input_sentence_size=10000000
--character_coverage=0.99995 --model_type=unigram
Owner
Google Research
Google Research
A complete speech segmentation system using Kaldi and x-vectors for voice activity detection (VAD) and speaker diarisation.

bbc-speech-segmenter: Voice Activity Detection & Speaker Diarization A complete speech segmentation system using Kaldi and x-vectors for voice activit

BBC 16 Oct 27, 2022
Fully Convlutional Neural Networks for state-of-the-art time series classification

Deep Learning for Time Series Classification As the simplest type of time series data, univariate time series provides a reasonably good starting poin

Stephen 572 Dec 23, 2022
Reference implementation of code generation projects from Facebook AI Research. General toolkit to apply machine learning to code, from dataset creation to model training and evaluation. Comes with pretrained models.

This repository is a toolkit to do machine learning for programming languages. It implements tokenization, dataset preprocessing, model training and m

Facebook Research 408 Jan 01, 2023
利用Tensorflow实现基于CNN的中文短文本分类

Text Classification with CNN 使用卷积神经网络进行中文文本分类 CNN做句子分类的论文可以参看: Convolutional Neural Networks for Sentence Classification 还可以去读dennybritz大牛的博客:Implemen

Jeremiah 4 Nov 08, 2022
An auto discord account and token generator. Automatically verifies the phone number. Works without proxy. Bypasses captcha.

JOIN DISCORD SERVER https://discord.gg/uAc3agBY FREE HCAPTCHA SOLVING API Discord-Token-Gen An auto discord token generator. Auto verifies phone numbe

3kp 271 Jan 01, 2023
AdamW optimizer and cosine learning rate annealing with restarts

AdamW optimizer and cosine learning rate annealing with restarts This repository contains an implementation of AdamW optimization algorithm and cosine

Maksym Pyrozhok 133 Dec 20, 2022
Learning to Prompt for Continual Learning

Learning to Prompt for Continual Learning (L2P) Official Jax Implementation L2P is a novel continual learning technique which learns to dynamically pr

Google Research 207 Jan 06, 2023
C3DPO - Canonical 3D Pose Networks for Non-rigid Structure From Motion.

C3DPO: Canonical 3D Pose Networks for Non-Rigid Structure From Motion By: David Novotny, Nikhila Ravi, Benjamin Graham, Natalia Neverova, Andrea Vedal

Meta Research 309 Dec 16, 2022
Vehicles Counting using YOLOv4 + DeepSORT + Flask + Ngrok

A project for counting vehicles using YOLOv4 + DeepSORT + Flask + Ngrok

Duong Tran Thanh 37 Dec 16, 2022
GUPNet - Geometry Uncertainty Projection Network for Monocular 3D Object Detection

GUPNet This is the official implementation of "Geometry Uncertainty Projection Network for Monocular 3D Object Detection". citation If you find our wo

Yan Lu 103 Dec 28, 2022
OCR Streamlit App is used to extract text from images using python's easyocr, pytorch and streamlit packages

OCR-Streamlit-App OCR Streamlit App is used to extract text from images using python's easyocr, pytorch and streamlit packages OCR app gets an image a

Siva Prakash 5 Apr 05, 2022
Context Axial Reverse Attention Network for Small Medical Objects Segmentation

CaraNet: Context Axial Reverse Attention Network for Small Medical Objects Segmentation This repository contains the implementation of a novel attenti

401 Dec 23, 2022
Cancer Drug Response Prediction via a Hybrid Graph Convolutional Network

DeepCDR Cancer Drug Response Prediction via a Hybrid Graph Convolutional Network This work has been accepted to ECCB2020 and was also published in the

Qiao Liu 50 Dec 18, 2022
This repository holds code and data for our PETS'22 article 'From "Onion Not Found" to Guard Discovery'.

From "Onion Not Found" to Guard Discovery (PETS'22) This repository holds the code and data for our PETS'22 paper titled 'From "Onion Not Found" to Gu

Lennart Oldenburg 3 May 04, 2022
A custom DeepStack model that has been trained detecting ONLY the USPS logo

This repository provides a custom DeepStack model that has been trained detecting ONLY the USPS logo. This was created after I discovered that the Deepstack OpenLogo custom model I was using did not

Stephen Stratoti 9 Dec 27, 2022
PyMove is a Python library to simplify queries and visualization of trajectories and other spatial-temporal data

Use PyMove and go much further Information Package Status License Python Version Platforms Build Status PyPi version PyPi Downloads Conda version Cond

Insight Data Science Lab 64 Nov 15, 2022
Detection of PCBA defect

Detection_of_PCBA_defect Detection_of_PCBA_defect Use yolov5 to train. $pip install -r requirements.txt Detect.py will detect file(jpg,mp4...) in cu

6 Nov 28, 2022
LaneDet is an open source lane detection toolbox based on PyTorch that aims to pull together a wide variety of state-of-the-art lane detection models

LaneDet is an open source lane detection toolbox based on PyTorch that aims to pull together a wide variety of state-of-the-art lane detection models. Developers can reproduce these SOTA methods and

TuZheng 405 Jan 04, 2023
Multi-Objective Loss Balancing for Physics-Informed Deep Learning

Multi-Objective Loss Balancing for Physics-Informed Deep Learning Code for ReLoBRaLo. Abstract Physics Informed Neural Networks (PINN) are algorithms

Rafael Bischof 16 Dec 12, 2022
Simple sinc interpolation in PyTorch.

Kazane: simple sinc interpolation for 1D signal in PyTorch Kazane utilize FFT based convolution to provide fast sinc interpolation for 1D signal when

Chin-Yun Yu 10 May 03, 2022