Train 🤗transformers with DeepSpeed: ZeRO-2, ZeRO-3

Overview

Fork from https://github.com/huggingface/transformers/tree/86d5fb0b360e68de46d40265e7c707fe68c8015b/examples/pytorch/language-modeling at 2021.05.17.

Language model training

Fine-tuning (or training from scratch) the library models for language modeling on a text dataset for GPT, GPT-2, ALBERT, BERT, DistilBERT, RoBERTa, XLNet... GPT and GPT-2 are trained or fine-tuned using a causal language modeling (CLM) loss while ALBERT, BERT, DistilBERT and RoBERTa are trained or fine-tuned using a masked language modeling (MLM) loss. XLNet uses permutation language modeling (PLM), you can find more information about the differences between those objectives in our model summary.

There are two sets of scripts provided. The first set leverages the Trainer API. The second set with no_trainer in the suffix uses a custom training loop and leverages the 🤗 Accelerate library . Both sets use the 🤗 Datasets library. You can easily customize them to your needs if you need extra processing on your datasets.

Note: The old script run_language_modeling.py is still available here.

The following examples, will run on datasets hosted on our hub or with your own text files for training and validation. We give examples of both below.

GPT-2/GPT and causal language modeling

The following example fine-tunes GPT-2 on WikiText-2. We're using the raw WikiText-2 (no tokens were replaced before the tokenization). The loss here is that of causal language modeling.

python run_clm.py \
    --model_name_or_path gpt2 \
    --dataset_name wikitext \
    --dataset_config_name wikitext-2-raw-v1 \
    --do_train \
    --do_eval \
    --output_dir /tmp/test-clm

This takes about half an hour to train on a single K80 GPU and about one minute for the evaluation to run. It reaches a score of ~20 perplexity once fine-tuned on the dataset.

To run on your own training and validation files, use the following command:

python run_clm.py \
    --model_name_or_path gpt2 \
    --train_file path_to_train_file \
    --validation_file path_to_validation_file \
    --do_train \
    --do_eval \
    --output_dir /tmp/test-clm

This uses the built in HuggingFace Trainer for training. If you want to use a custom training loop, you can utilize or adapt the run_clm_no_trainer.py script. Take a look at the script for a list of supported arguments. An example is shown below:

python run_clm_no_trainer.py \
    --dataset_name wikitext \
    --dataset_config_name wikitext-2-raw-v1 \
    --model_name_or_path gpt2 \
    --output_dir /tmp/test-clm

RoBERTa/BERT/DistilBERT and masked language modeling

The following example fine-tunes RoBERTa on WikiText-2. Here too, we're using the raw WikiText-2. The loss is different as BERT/RoBERTa have a bidirectional mechanism; we're therefore using the same loss that was used during their pre-training: masked language modeling.

In accordance to the RoBERTa paper, we use dynamic masking rather than static masking. The model may, therefore, converge slightly slower (over-fitting takes more epochs).

python run_mlm.py \
    --model_name_or_path roberta-base \
    --dataset_name wikitext \
    --dataset_config_name wikitext-2-raw-v1 \
    --do_train \
    --do_eval \
    --output_dir /tmp/test-mlm

To run on your own training and validation files, use the following command:

python run_mlm.py \
    --model_name_or_path roberta-base \
    --train_file path_to_train_file \
    --validation_file path_to_validation_file \
    --do_train \
    --do_eval \
    --output_dir /tmp/test-mlm

If your dataset is organized with one sample per line, you can use the --line_by_line flag (otherwise the script concatenates all texts and then splits them in blocks of the same length).

This uses the built in HuggingFace Trainer for training. If you want to use a custom training loop, you can utilize or adapt the run_mlm_no_trainer.py script. Take a look at the script for a list of supported arguments. An example is shown below:

python run_mlm_no_trainer.py \
    --dataset_name wikitext \
    --dataset_config_name wikitext-2-raw-v1 \
    --model_name_or_path roberta-base \
    --output_dir /tmp/test-mlm

Note: On TPU, you should use the flag --pad_to_max_length in conjunction with the --line_by_line flag to make sure all your batches have the same length.

Whole word masking

This part was moved to examples/research_projects/mlm_wwm.

XLNet and permutation language modeling

XLNet uses a different training objective, which is permutation language modeling. It is an autoregressive method to learn bidirectional contexts by maximizing the expected likelihood over all permutations of the input sequence factorization order.

We use the --plm_probability flag to define the ratio of length of a span of masked tokens to surrounding context length for permutation language modeling.

The --max_span_length flag may also be used to limit the length of a span of masked tokens used for permutation language modeling.

Here is how to fine-tune XLNet on wikitext-2:

python run_plm.py \
    --model_name_or_path=xlnet-base-cased \
    --dataset_name wikitext \
    --dataset_config_name wikitext-2-raw-v1 \
    --do_train \
    --do_eval \
    --output_dir /tmp/test-plm

To fine-tune it on your own training and validation file, run:

python run_plm.py \
    --model_name_or_path=xlnet-base-cased \
    --train_file path_to_train_file \
    --validation_file path_to_validation_file \
    --do_train \
    --do_eval \
    --output_dir /tmp/test-plm

If your dataset is organized with one sample per line, you can use the --line_by_line flag (otherwise the script concatenates all texts and then splits them in blocks of the same length).

Note: On TPU, you should use the flag --pad_to_max_length in conjunction with the --line_by_line flag to make sure all your batches have the same length.

Owner
Junbum Lee
AI/NLP Researcher who loves python, web, and data. And who believes technology would make the world a better place.
Junbum Lee
A spaCy wrapper of OpenTapioca for named entity linking on Wikidata

spaCyOpenTapioca A spaCy wrapper of OpenTapioca for named entity linking on Wikidata. Table of contents Installation How to use Local OpenTapioca Vizu

Universitätsbibliothek Mannheim 80 Jan 03, 2023
Bidirectional Variational Inference for Non-Autoregressive Text-to-Speech (BVAE-TTS)

Bidirectional Variational Inference for Non-Autoregressive Text-to-Speech (BVAE-TTS) Yoonhyung Lee, Joongbo Shin, Kyomin Jung Abstract: Although early

LEE YOON HYUNG 147 Dec 05, 2022
The PyTorch based implementation of continuous integrate-and-fire (CIF) module.

CIF-PyTorch This is a PyTorch based implementation of continuous integrate-and-fire (CIF) module for end-to-end (E2E) automatic speech recognition (AS

Minglun Han 24 Dec 29, 2022
Two-stage text summarization with BERT and BART

Two-Stage Text Summarization Description We experiment with a 2-stage summarization model on CNN/DailyMail dataset that combines the ability to filter

Yukai Yang (Alexis) 6 Oct 22, 2022
Machine learning classifiers to predict American Sign Language .

ASL-Classifiers American Sign Language (ASL) is a natural language that serves as the predominant sign language of Deaf communities in the United Stat

Tarek idrees 0 Feb 08, 2022
OpenChat: Opensource chatting framework for generative models

OpenChat is opensource chatting framework for generative models.

Hyunwoong Ko 427 Jan 06, 2023
Sequence modeling benchmarks and temporal convolutional networks

Sequence Modeling Benchmarks and Temporal Convolutional Networks (TCN) This repository contains the experiments done in the work An Empirical Evaluati

CMU Locus Lab 3.5k Jan 03, 2023
Officile code repository for "A Game-Theoretic Perspective on Risk-Sensitive Reinforcement Learning"

CvarAdversarialRL Official code repository for "A Game-Theoretic Perspective on Risk-Sensitive Reinforcement Learning". Initial setup Create a virtual

Mathieu Godbout 1 Nov 19, 2021
Code for our ACL 2021 paper - ConSERT: A Contrastive Framework for Self-Supervised Sentence Representation Transfer

ConSERT Code for our ACL 2021 paper - ConSERT: A Contrastive Framework for Self-Supervised Sentence Representation Transfer Requirements torch==1.6.0

Yan Yuanmeng 478 Dec 25, 2022
Repository for Graph2Pix: A Graph-Based Image to Image Translation Framework

Graph2Pix: A Graph-Based Image to Image Translation Framework Installation Install the dependencies in env.yml $ conda env create -f env.yml $ conda a

18 Nov 17, 2022
Fidibo.com comments Sentiment Analyser

Fidibo.com comments Sentiment Analyser Introduction This project first asynchronously grab Fidibo.com books comment data using grabber.py and then sav

Iman Kermani 3 Apr 15, 2022
AudioCLIP Extending CLIP to Image, Text and Audio

AudioCLIP Extending CLIP to Image, Text and Audio This repository contains implementation of the models described in the paper arXiv:2106.13043. This

458 Jan 02, 2023
Official code repository of the paper Linear Transformers Are Secretly Fast Weight Programmers.

Linear Transformers Are Secretly Fast Weight Programmers This repository contains the code accompanying the paper Linear Transformers Are Secretly Fas

Imanol Schlag 77 Dec 19, 2022
ProteinBERT is a universal protein language model pretrained on ~106M proteins from the UniRef90 dataset.

ProteinBERT is a universal protein language model pretrained on ~106M proteins from the UniRef90 dataset. Through its Python API, the pretrained model can be fine-tuned on any protein-related task in

241 Jan 04, 2023
⚡ boost inference speed of T5 models by 5x & reduce the model size by 3x using fastT5.

Reduce T5 model size by 3X and increase the inference speed up to 5X. Install Usage Details Functionalities Benchmarks Onnx model Quantized onnx model

Kiran R 399 Jan 05, 2023
A PyTorch implementation of paper "Learning Shared Semantic Space for Speech-to-Text Translation", ACL (Findings) 2021

Chimera: Learning Shared Semantic Space for Speech-to-Text Translation This is a Pytorch implementation for the "Chimera" paper Learning Shared Semant

Chi Han 43 Dec 28, 2022
ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators

ELECTRA Introduction ELECTRA is a method for self-supervised language representation learning. It can be used to pre-train transformer networks using

Google Research 2.1k Dec 28, 2022
This is Assignment1 code for the Web Data Processing System.

This is a Python program to Entity Linking by processing WARC files. We recognize entities from web pages and link them to a Knowledge Base(Wikidata).

3 Dec 04, 2022
A minimal Conformer ASR implementation adapted from ESPnet.

Conformer ASR A minimal Conformer ASR implementation adapted from ESPnet. Introduction I want to use the pre-trained English ASR model provided by ESP

Niu Zhe 3 Jan 24, 2022
CCF BDCI 2020 房产行业聊天问答匹配赛道 A榜47/2985

CCF BDCI 2020 房产行业聊天问答匹配 A榜47/2985 赛题描述详见:https://www.datafountain.cn/competitions/474 文件说明 data: 存放训练数据和测试数据以及预处理代码 model_bert.py: 网络模型结构定义 adv_train

shuo 40 Sep 28, 2022