Merge branch 'hepj-test' into 'main'

更新transformer代码

See merge request dcutoolkit/deeplearing/dlexamples_new!47

PyTorch/NLP/new-Transformer/README.md

+### 1.Benchmark下载地址
+
+transformer：https://github.com/pytorch/fairseq.git
+
+最新的githup上代码和现在存在区别，最好使用本地旧版本代码，否则可能出现HIP版本的torch不识别情况
+
+### 2.1 数据集准备
+
+```
+环境准备：
+
+pip3 install fastBPE sacremoses subword_nmt
+
+数据集下载：
+
+cd examples/translation/
+
+bash prepare-wmt14en2de.sh --icml17
+
+数据预处理：
+
+DATA_DIR=~/data/wmt14_en_de_joined_dict
+
+TEXT=`pwd`/examples/translation/wmt14_en_de
+
+fairseq-preprocess --source-lang en --target-lang de --trainpref $TEXT/train --validpref $TEXT/valid --testpref $TEXT/test --destdir $DATA_DIR --nwordssrc 32768 --nwordstgt 32768 --joined-dictionary --workers 20
+
+相关参数说明：
+
+--source-lang   source language
+
+--target-lang   target language
+
+--trainpref     rain file prefix (also used to build dictionaries)
+
+--validpref     comma separated, valid file prefixes (words missing from train set are replaced with <unk>)
+
+--testpref      comma separated, test file prefixes (words missing from train set are replaced with <unk>)
+
+--destdir       destination dir, Default: “data-bin”
+
+数据集路径：
+
+DATA_PATH=~/data/wmt14_en_de_joined_dict
+```
+
+
+
+#### 2.2.环境部署
+
+##### 2.2.1.构建测试的虚拟环境
+
+```
+virtualenv -p python3 venv
+
+source venv/bin/activate
+```
+
+##### 2.2.2.安装python3.7环境下的依赖包
+
+```
+pip3 install --upgrade pip
+
+pip3 install typing
+
+pip3 install sacremoses
+
+pip3 install numpy
+
+pip3 install torch-1.10.0a0+gitd8cde89.atomic.dtk22042-cp37-cp37m-linux_x86_64.whl
+
+pip3 install apex-0.1_dtk22.04-cp37-cp37m-linux_x86_64.whl
+
+pip3 install setuptools==59.5.0
+
+pip3 install protobuf==3.20.0
+```
+
+##### 2.2.3.安装fairseq
+
+```
+#git clone https://github.com/pytorch/fairseq.git
+#cd fairseq
+#整个transformer文件夹就是从github上拷贝下来的fariseq，可以直接安装
+pip3 install --editable ./
+
+相关说明：可以先把setup.py里的torch、torch-audio的安装屏蔽掉
+```
+
+##### 2.2.4.环境变量里设置env.sh
+
+```
+WORK_PATH=`pwd`
+source env.sh
+```
+
+### 3.transformer测试（昆山）
+
+#### 3.1.单卡测试（单精度）
+
+##### 3.1.1.run.sh
+
+```
+
+export HSA_FORCE_FINE_GRAIN_PCIE=1
+export MIOPEN_FIND_MODE=1
+export HIP_VISIBLE_DEVICES=0
+export TOKEN=2560
+export DATA_PATH=~/data/wmt14_en_de_joined_dict
+
+export HIP_LAUNCH_BLOCKING=1
+export ROCBLAS_ATOMICS_MOD=1
+python3 train.py \
+    $DATA_PATH \
+    --arch transformer_wmt_en_de \
+    --share-decoder-input-output-embed \
+    --optimizer adam \
+    --adam-betas '(0.9, 0.98)' \
+    --clip-norm 0.0 \
+    --lr 5e-4 \
+    --lr-scheduler inverse_sqrt \
+    --warmup-updates 4000 \
+    --dropout 0.3 \
+    --weight-decay 0.0001 \
+    --criterion label_smoothed_cross_entropy \
+    --label-smoothing 0.1 \
+    --max-tokens ${TOKEN} \
+    --eval-bleu \
+    --eval-bleu-args '{"beam": 5, "max_len_a": 1.2, "max_len_b": 10}' \
+    --eval-bleu-detok moses \
+    --eval-bleu-remove-bpe \
+    --eval-bleu-print-samples \
+    --best-checkpoint-metric bleu \
+    --maximize-best-checkpoint-metric \
+    --max-epoch 1
+
+```
+
+##### 3.1.2.运行
+```
+./ run.sh
+```
+
+#### 3.2.四卡测试（单精度）
+
+##### 3.2.1.single_process.sh
+
+```
+#!/bin/bash
+
+export MIOPEN_DEBUG_DISABLE_FIND_DB=1
+
+export NCCL_SOCKET_IFNAME=eno1
+
+export HSA_USERPTR_FOR_PAGED_MEM=0
+
+lrank=$OMPI_COMM_WORLD_LOCAL_RANK
+
+comm_rank=$OMPI_COMM_WORLD_RANK
+
+comm_size=$OMPI_COMM_WORLD_SIZE
+
+TOKENS=2560
+
+DATA_PATH=~/fairseq/examples/translation/data-bin/wmt14_en_de_joined_dict
+
+APP="python3 ~/fairseq/train.py $DATA_PATH --arch transformer_wmt_en_de --share-decoder-input-output-embed --optimizer adam --adam-betas (0.9,0.98) --clip-norm 0.0 --lr 5e-4 --lr-scheduler inverse_sqrt --warmup-updates 4000 --dropout 0.3 --weight-decay 0.0001 --criterion label_smoothed_cross_entropy --label-smoothing 0.1 --max-tokens $TOKENS --eval-bleu --eval-bleu-args {\"beam\":5,\"max_len_a\":1.2,\"max_len_b\":10} --eval-bleu-detok moses --eval-bleu-remove-bpe --eval-bleu-print-samples --best-checkpoint-metric bleu --maximize-best-checkpoint-metric --distributed-rank ${comm_rank} --distributed-world-size ${comm_size} --device-id ${lrank} --local_rank ${lrank} --distributed-init-method tcp://${1}:34567 --distributed-no-spawn --max-epoch 1"
+
+case ${lrank} in
+
+[0])
+
+  export HIP_VISIBLE_DEVICES=0,1,2,3
+
+  export UCX_NET_DEVICES=mlx5_0:1
+
+  export UCX_IB_PCI_BW=mlx5_0:50Gbs
+
+  echo NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=0 --membind=0 ${APP}
+
+  NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=0 --membind=0 ${APP}
+
+  ;;
+
+[1])
+
+  export HIP_VISIBLE_DEVICES=0,1,2,3
+
+  export UCX_NET_DEVICES=mlx5_1:1
+
+  export UCX_IB_PCI_BW=mlx5_1:50Gbs
+
+  echo NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=1 --membind=1 ${APP}
+
+  NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=1 --membind=1 ${APP}
+
+  ;;
+
+[2])
+
+  export HIP_VISIBLE_DEVICES=0,1,2,3
+
+  export UCX_NET_DEVICES=mlx5_2:1
+
+  export UCX_IB_PCI_BW=mlx5_2:50Gbs
+
+  echo NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=2 --membind=2 ${APP}
+
+  NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=2 --membind=2 ${APP}
+
+  ;;
+
+[3])
+
+  export HIP_VISIBLE_DEVICES=0,1,2,3
+
+  export UCX_NET_DEVICES=mlx5_3:1
+
+  export UCX_IB_PCI_BW=mlx5_3:50Gbs
+
+  echo NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=3 --membind=3 ${APP}
+
+  NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=3 --membind=3 ${APP}
+
+  ;;
+
+esac
+```
+
+##### 3.2.2.run4.sh
+
+```
+#!/usr/bin/env bash
+#SBATCH -J distribute
+#SBATCH -p wzhdtest
+#SBATCH -N 1
+#SBARCH -n 32
+#SBATCH --ntasks-per-node=4
+#SBATCH --cpus-per-task=8
+#SBATCH --gres=dcu:4
+set -x
+hostfile=./$SLURM_JOB_ID
+scontrol show hostnames $SLURM_JOB_NODELIST > ${hostfile}
+for i in `cat $hostfile`
+do
+    echo ${i} slots=4 >> `pwd`/hostfile-$SLURM_JOB_ID
+    ((num_node=${num_node}+1))
+done
+num_dcu=$((${num_node}*4))
+echo $num_dcu
+nodename=$(cat $hostfile |sed -n "1p")
+echo $nodename
+dist_url=`echo $nodename | awk '{print $1}'`
+export HSA_USERPTR_FOR_PAGED_MEM=0
+mpirun -np ${num_dcu} --hostfile hostfile-$SLURM_JOB_ID single_process.sh  $dist_url
+```
+
+##### 3.2.3.运行
+```
+sbatch run4.sh
+```
+
+##### 3.2.4.参数说明
+
+- 上面的中single_process.sh需要关注--max-tokens；
+-  通过--arch 设置要测试的网络，eg:transformer_wmt_en_de 等；
+- 上述 run_transformer_4dcus.sh中mpirun 运行命令表示使用4张DCU加速卡训练。
+
+#### 3.3.单卡测试（半精度）
+
+##### 3.3.1.fp16_run_transformer.sh
+
+```
+export HSA_FORCE_FINE_GRAIN_PCIE=1
+export MIOPEN_FIND_MODE=1
+export HIP_VISIBLE_DEVICES=0
+export DATA_PATH=~/data/wmt14_en_de_joined_dict
+export TOKEN=2560
+python3 train.py \
+    $DATA_PATH \
+    --save-dir module-fp16_2560 \
+    --arch transformer_wmt_en_de --share-decoder-input-output-embed \
+    --optimizer adam --adam-betas '(0.9, 0.98)' --clip-norm 0.0 \
+    --lr 5e-4 --lr-scheduler inverse_sqrt --warmup-updates 4000 \
+    --dropout 0.3 --weight-decay 0.0001 \
+    --criterion label_smoothed_cross_entropy --label-smoothing 0.1 \
+    --max-tokens ${TOKEN} \
+    --eval-bleu \
+    --eval-bleu-args '{"beam": 5, "max_len_a": 1.2, "max_len_b": 10}' \
+    --eval-bleu-detok moses \
+    --eval-bleu-remove-bpe \
+    --eval-bleu-print-samples \
+    --best-checkpoint-metric bleu --maximize-best-checkpoint-metric --max-epoch 1 --fp16
+```
+
+##### 3.3.2.运行
+
+./ fp16_run_transformer.sh
+
+##### 3.3.3.参数说明
+
+--max-tokens 根据tokens设置batch size
+
+--fp16  使用半精度训练
+
+#### 3.4.四卡测试（半精度）
+
+##### 3.4.1.fp16_single_process.sh
+
+```
+
+export HIP_VISIBLE_DEVICES=0
+export MIOPEN_DEBUG_DISABLE_FIND_DB=1
+export NCCL_SOCKET_IFNAME=eno1
+export HSA_USERPTR_FOR_PAGED_MEM=0
+lrank=$OMPI_COMM_WORLD_LOCAL_RANK
+comm_rank=$OMPI_COMM_WORLD_RANK
+comm_size=$OMPI_COMM_WORLD_SIZE
+TOKENS=2560
+DATA_PATH=~/data/wmt14_en_de_joined_dict
+APP="python3 ~/fairseq/train.py $DATA_PATH --arch transformer_wmt_en_de --share-decoder-input-output-embed --optimizer adam --adam-betas '(0.9,0.98)' --clip-norm 0.0 --lr 5e-4 --lr-scheduler inverse_sqrt --warmup-updates 4000 --dropout 0.3 --weight-decay 0.0001 --criterion label_smoothed_cross_entropy --label-smoothing 0.1 --max-tokens $TOKENS --eval-bleu --eval-bleu-args {\"beam\":5,\"max_len_a\":1.2,\"max_len_b\":10} --eval-bleu-detok moses --eval-bleu-remove-bpe --eval-bleu-print-samples --best-checkpoint-metric bleu --maximize-best-checkpoint-metric --distributed-rank ${comm_rank} --distributed-world-size ${comm_size} --device-id ${lrank} --local_rank ${lrank} --distributed-init-method tcp://${1}:34567 --distributed-no-spawn --max-epoch 1 --fp16"
+case ${lrank} in
+[0])
+  export HIP_VISIBLE_DEVICES=0,1,2,3
+  export UCX_NET_DEVICES=mlx5_0:1
+  export UCX_IB_PCI_BW=mlx5_0:50Gbs
+  echo NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=0 --membind=0 ${APP}
+  NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=0 --membind=0 ${APP}
+  ;;
+[1])
+  export HIP_VISIBLE_DEVICES=0,1,2,3
+  export UCX_NET_DEVICES=mlx5_1:1
+  export UCX_IB_PCI_BW=mlx5_1:50Gbs
+  echo NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=1 --membind=1 ${APP}
+  NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=1 --membind=1 ${APP}
+  ;;
+[2])
+  export HIP_VISIBLE_DEVICES=0,1,2,3
+  export UCX_NET_DEVICES=mlx5_2:1
+  export UCX_IB_PCI_BW=mlx5_2:50Gbs
+  echo NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=2 --membind=2 ${APP} 
+  NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=2 --membind=2 ${APP}
+  ;;
+[3])
+  export HIP_VISIBLE_DEVICES=0,1,2,3
+  export UCX_NET_DEVICES=mlx5_3:1
+  export UCX_IB_PCI_BW=mlx5_3:50Gbs
+  echo NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=3 --membind=3 ${APP}
+  NCCL_SOCKET_IFNAME=ib0 numactl --cpunodebind=3 --membind=3 ${APP}
+  ;;
+esac
+```
+
+##### 3.4.2.fp16_run_transformer_4dcus.sh
+
+```
+#!/usr/bin/env bash
+#SBATCH -J transformer
+#SBATCH -p wzhdtest
+#SBATCH -N 1
+#SBARCH -n 32
+#SBATCH --ntasks-per-node=4
+#SBATCH --cpus-per-task=8
+#SBATCH --gres=dcu:4
+set -x
+hostfile=./$SLURM_JOB_ID
+scontrol show hostnames $SLURM_JOB_NODELIST > ${hostfile}
+for i in `cat $hostfile`
+do
+    echo ${i} slots=4 >> `pwd`/hostfile-$SLURM_JOB_ID
+    ((num_node=${num_node}+1))
+done
+num_dcu=$((${num_node}*4))
+echo $num_dcu
+nodename=$(cat $hostfile |sed -n "1p")
+echo $nodename
+dist_url=`echo $nodename | awk '{print $1}'`
+export HSA_USERPTR_FOR_PAGED_MEM=0
+mpirun -np ${num_dcu} --hostfile hostfile-$SLURM_JOB_ID single_fp16.sh  $dist_ur
+
+```
+
+##### 3.4.3.运行
+
+sbatch fp16_ run_transformer_4dcus.sh
+
+##### 3.4.4.参数说明
+
+- 上面的中single_process.sh需要关注--max-tokens；
+-  通过--arch 设置要测试的网络，eg:transformer_wmt_en_de 等；
+- 上述 run_transformer_4dcus.sh中mpirun 运行命令表示使用4张DCU加速卡训练。
+
+#### 3.5. 部分问题说明
+
+##### 3.5.1. format错误
+
+报错信息如下：
+
+```
+  File "~/virturlenv-test/venv/lib/python3.6/site-packages/sacrebleu/metrics/bleu.py", line 103, in __init__
+    self._verbose += f"ratio = {self.ratio:.3f} hyp_len = {self.sys_len:d} "
+ ...
+
+ValueError: Unknown format code 'd' for object of type 'float'
+```
+
+修改方法：修改报错提示中的bleu.py，103和104行的d改成.0f
+
+```
+#修改后
+self._verbose += f"ratio = {self.ratio:.3f} hyp_len ={self.sys_len:.0f}"
+self._verbose += f"ref_len = {slef.ref_len:.0f}"
+```
+
+
+
+##### 3.5.2 json格式解析错误
+
+报错信息如下：
+
+json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)
+
+解析错误的地方，可以看出是引号嵌套问题
+
+错误的打印
+
+```
+'eval_bleu_args': "'{beam:5,max_len_a:1.2,max_len_b:10}'
+```
+
+正确的打印
+
+```
+'eval_bleu_args': '{"beam": 5, "max_len_a": 1.2, "max_len_b": 10}'
+```
+
+修改为：
+
+```
+" --eval-bleu-args {\"beam\":5,\"max_len_a\":1.2,\"max_len_b\":10}"
+```
+

PyTorch/NLP/new-Transformer/README.md-old

+<p align="center">
+  <img src="docs/fairseq_logo.png" width="150">
+  <br />
+  <br />
+  <a href="https://opensource.fb.com/support-ukraine"><img alt="Support Ukraine" src="https://img.shields.io/badge/Support-Ukraine-FFD500?style=flat&labelColor=005BBB" /></a>
+  <a href="https://github.com/pytorch/fairseq/blob/main/LICENSE"><img alt="MIT License" src="https://img.shields.io/badge/license-MIT-blue.svg" /></a>
+  <a href="https://github.com/pytorch/fairseq/releases"><img alt="Latest Release" src="https://img.shields.io/github/release/pytorch/fairseq.svg" /></a>
+  <a href="https://github.com/pytorch/fairseq/actions?query=workflow:build"><img alt="Build Status" src="https://github.com/pytorch/fairseq/workflows/build/badge.svg" /></a>
+  <a href="https://fairseq.readthedocs.io/en/latest/?badge=latest"><img alt="Documentation Status" src="https://readthedocs.org/projects/fairseq/badge/?version=latest" /></a>
+</p>
+
+--------------------------------------------------------------------------------
+
+Fairseq(-py) is a sequence modeling toolkit that allows researchers and
+developers to train custom models for translation, summarization, language
+modeling and other text generation tasks.
+
+We provide reference implementations of various sequence modeling papers:
+
+<details><summary>List of implemented papers</summary><p>
+
+* **Convolutional Neural Networks (CNN)**
+  + [Language Modeling with Gated Convolutional Networks (Dauphin et al., 2017)](examples/language_model/conv_lm/README.md)
+  + [Convolutional Sequence to Sequence Learning (Gehring et al., 2017)](examples/conv_seq2seq/README.md)
+  + [Classical Structured Prediction Losses for Sequence to Sequence Learning (Edunov et al., 2018)](https://github.com/pytorch/fairseq/tree/classic_seqlevel)
+  + [Hierarchical Neural Story Generation (Fan et al., 2018)](examples/stories/README.md)
+  + [wav2vec: Unsupervised Pre-training for Speech Recognition (Schneider et al., 2019)](examples/wav2vec/README.md)
+* **LightConv and DynamicConv models**
+  + [Pay Less Attention with Lightweight and Dynamic Convolutions (Wu et al., 2019)](examples/pay_less_attention_paper/README.md)
+* **Long Short-Term Memory (LSTM) networks**
+  + Effective Approaches to Attention-based Neural Machine Translation (Luong et al., 2015)
+* **Transformer (self-attention) networks**
+  + Attention Is All You Need (Vaswani et al., 2017)
+  + [Scaling Neural Machine Translation (Ott et al., 2018)](examples/scaling_nmt/README.md)
+  + [Understanding Back-Translation at Scale (Edunov et al., 2018)](examples/backtranslation/README.md)
+  + [Adaptive Input Representations for Neural Language Modeling (Baevski and Auli, 2018)](examples/language_model/README.adaptive_inputs.md)
+  + [Lexically constrained decoding with dynamic beam allocation (Post & Vilar, 2018)](examples/constrained_decoding/README.md)
+  + [Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context (Dai et al., 2019)](examples/truncated_bptt/README.md)
+  + [Adaptive Attention Span in Transformers (Sukhbaatar et al., 2019)](examples/adaptive_span/README.md)
+  + [Mixture Models for Diverse Machine Translation: Tricks of the Trade (Shen et al., 2019)](examples/translation_moe/README.md)
+  + [RoBERTa: A Robustly Optimized BERT Pretraining Approach (Liu et al., 2019)](examples/roberta/README.md)
+  + [Facebook FAIR's WMT19 News Translation Task Submission (Ng et al., 2019)](examples/wmt19/README.md)
+  + [Jointly Learning to Align and Translate with Transformer Models (Garg et al., 2019)](examples/joint_alignment_translation/README.md )
+  + [Multilingual Denoising Pre-training for Neural Machine Translation (Liu et at., 2020)](examples/mbart/README.md)
+  + [Neural Machine Translation with Byte-Level Subwords (Wang et al., 2020)](examples/byte_level_bpe/README.md)
+  + [Unsupervised Quality Estimation for Neural Machine Translation (Fomicheva et al., 2020)](examples/unsupervised_quality_estimation/README.md)
+  + [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations (Baevski et al., 2020)](examples/wav2vec/README.md)
+  + [Generating Medical Reports from Patient-Doctor Conversations Using Sequence-to-Sequence Models (Enarvi et al., 2020)](examples/pointer_generator/README.md)
+  + [Linformer: Self-Attention with Linear Complexity (Wang et al., 2020)](examples/linformer/README.md)
+  + [Cross-lingual Retrieval for Iterative Self-Supervised Training (Tran et al., 2020)](examples/criss/README.md)
+  + [Deep Transformers with Latent Depth (Li et al., 2020)](examples/latent_depth/README.md)
+  + [Unsupervised Cross-lingual Representation Learning for Speech Recognition (Conneau et al., 2020)](https://arxiv.org/abs/2006.13979)
+  + [Self-training and Pre-training are Complementary for Speech Recognition (Xu et al., 2020)](https://arxiv.org/abs/2010.11430)
+  + [Robust wav2vec 2.0: Analyzing Domain Shift in Self-Supervised Pre-Training (Hsu, et al., 2021)](https://arxiv.org/abs/2104.01027)
+  + [Unsupervised Speech Recognition (Baevski, et al., 2021)](https://arxiv.org/abs/2105.11084)
+  + [Simple and Effective Zero-shot Cross-lingual Phoneme Recognition (Xu et al., 2021)](https://arxiv.org/abs/2109.11680)
+  + [VideoCLIP: Contrastive Pre-training for Zero-shot Video-Text Understanding (Xu et. al., 2021)](https://arxiv.org/pdf/2109.14084.pdf)
+  + [VLM: Task-agnostic Video-Language Model Pre-training for Video Understanding (Xu et. al., 2021)](https://aclanthology.org/2021.findings-acl.370.pdf)
+  + [NormFormer: Improved Transformer Pretraining with Extra Normalization (Shleifer et. al, 2021)](examples/normformer/README.md)
+* **Non-autoregressive Transformers**
+  + Non-Autoregressive Neural Machine Translation (Gu et al., 2017)
+  + Deterministic Non-Autoregressive Neural Sequence Modeling by Iterative Refinement (Lee et al. 2018)
+  + Insertion Transformer: Flexible Sequence Generation via Insertion Operations (Stern et al. 2019)
+  + Mask-Predict: Parallel Decoding of Conditional Masked Language Models (Ghazvininejad et al., 2019)
+  + [Levenshtein Transformer (Gu et al., 2019)](examples/nonautoregressive_translation/README.md)
+* **Finetuning**
+  + [Better Fine-Tuning by Reducing Representational Collapse (Aghajanyan et al. 2020)](examples/rxf/README.md)
+
+</p></details>
+
+### What's New:
+* May 2022 [Integration with xFormers](https://github.com/facebookresearch/xformers)
+* December 2021 [Released Direct speech-to-speech translation code](examples/speech_to_speech/README.md)
+* October 2021 [Released VideoCLIP and VLM models](examples/MMPT/README.md)
+* October 2021 [Released multilingual finetuned XLSR-53 model](examples/wav2vec/README.md)
+* September 2021 [`master` branch renamed to `main`](https://github.com/github/renaming).
+* July 2021 [Released DrNMT code](examples/discriminative_reranking_nmt/README.md)
+* July 2021 [Released Robust wav2vec 2.0 model](examples/wav2vec/README.md)
+* June 2021 [Released XLMR-XL and XLMR-XXL models](examples/xlmr/README.md)
+* May 2021 [Released Unsupervised Speech Recognition code](examples/wav2vec/unsupervised/README.md)
+* March 2021 [Added full parameter and optimizer state sharding + CPU offloading](examples/fully_sharded_data_parallel/README.md)
+* February 2021 [Added LASER training code](examples/laser/README.md)
+* December 2020: [Added Adaptive Attention Span code](examples/adaptive_span/README.md)
+* December 2020: [GottBERT model and code released](examples/gottbert/README.md)
+* November 2020: Adopted the [Hydra](https://github.com/facebookresearch/hydra) configuration framework
+  * [see documentation explaining how to use it for new and existing projects](docs/hydra_integration.md)
+* November 2020: [fairseq 0.10.0 released](https://github.com/pytorch/fairseq/releases/tag/v0.10.0)
+* October 2020: [Added R3F/R4F (Better Fine-Tuning) code](examples/rxf/README.md)
+* October 2020: [Deep Transformer with Latent Depth code released](examples/latent_depth/README.md)
+* October 2020: [Added CRISS models and code](examples/criss/README.md)
+
+<details><summary>Previous updates</summary><p>
+
+* September 2020: [Added Linformer code](examples/linformer/README.md)
+* September 2020: [Added pointer-generator networks](examples/pointer_generator/README.md)
+* August 2020: [Added lexically constrained decoding](examples/constrained_decoding/README.md)
+* August 2020: [wav2vec2 models and code released](examples/wav2vec/README.md)
+* July 2020: [Unsupervised Quality Estimation code released](examples/unsupervised_quality_estimation/README.md)
+* May 2020: [Follow fairseq on Twitter](https://twitter.com/fairseq)
+* April 2020: [Monotonic Multihead Attention code released](examples/simultaneous_translation/README.md)
+* April 2020: [Quant-Noise code released](examples/quant_noise/README.md)
+* April 2020: [Initial model parallel support and 11B parameters unidirectional LM released](examples/megatron_11b/README.md)
+* March 2020: [Byte-level BPE code released](examples/byte_level_bpe/README.md)
+* February 2020: [mBART model and code released](examples/mbart/README.md)
+* February 2020: [Added tutorial for back-translation](https://github.com/pytorch/fairseq/tree/main/examples/backtranslation#training-your-own-model-wmt18-english-german)
+* December 2019: [fairseq 0.9.0 released](https://github.com/pytorch/fairseq/releases/tag/v0.9.0)
+* November 2019: [VizSeq released (a visual analysis toolkit for evaluating fairseq models)](https://facebookresearch.github.io/vizseq/docs/getting_started/fairseq_example)
+* November 2019: [CamemBERT model and code released](examples/camembert/README.md)
+* November 2019: [BART model and code released](examples/bart/README.md)
+* November 2019: [XLM-R models and code released](examples/xlmr/README.md)
+* September 2019: [Nonautoregressive translation code released](examples/nonautoregressive_translation/README.md)
+* August 2019: [WMT'19 models released](examples/wmt19/README.md)
+* July 2019: fairseq relicensed under MIT license
+* July 2019: [RoBERTa models and code released](examples/roberta/README.md)
+* June 2019: [wav2vec models and code released](examples/wav2vec/README.md)
+
+</p></details>
+
+### Features:
+
+* multi-GPU training on one machine or across multiple machines (data and model parallel)
+* fast generation on both CPU and GPU with multiple search algorithms implemented:
+  + beam search
+  + Diverse Beam Search ([Vijayakumar et al., 2016](https://arxiv.org/abs/1610.02424))
+  + sampling (unconstrained, top-k and top-p/nucleus)
+  + [lexically constrained decoding](examples/constrained_decoding/README.md) (Post & Vilar, 2018)
+* [gradient accumulation](https://fairseq.readthedocs.io/en/latest/getting_started.html#large-mini-batch-training-with-delayed-updates) enables training with large mini-batches even on a single GPU
+* [mixed precision training](https://fairseq.readthedocs.io/en/latest/getting_started.html#training-with-half-precision-floating-point-fp16) (trains faster with less GPU memory on [NVIDIA tensor cores](https://developer.nvidia.com/tensor-cores))
+* [extensible](https://fairseq.readthedocs.io/en/latest/overview.html): easily register new models, criterions, tasks, optimizers and learning rate schedulers
+* [flexible configuration](docs/hydra_integration.md) based on [Hydra](https://github.com/facebookresearch/hydra) allowing a combination of code, command-line and file based configuration
+* [full parameter and optimizer state sharding](examples/fully_sharded_data_parallel/README.md)
+* [offloading parameters to CPU](examples/fully_sharded_data_parallel/README.md)
+
+We also provide [pre-trained models for translation and language modeling](#pre-trained-models-and-examples)
+with a convenient `torch.hub` interface:
+
+``` python
+en2de = torch.hub.load('pytorch/fairseq', 'transformer.wmt19.en-de.single_model')
+en2de.translate('Hello world', beam=5)
+# 'Hallo Welt'
+```
+
+See the PyTorch Hub tutorials for [translation](https://pytorch.org/hub/pytorch_fairseq_translation/)
+and [RoBERTa](https://pytorch.org/hub/pytorch_fairseq_roberta/) for more examples.
+
+# Requirements and Installation
+
+* [PyTorch](http://pytorch.org/) version >= 1.5.0
+* Python version >= 3.6
+* For training new models, you'll also need an NVIDIA GPU and [NCCL](https://github.com/NVIDIA/nccl)
+* **To install fairseq** and develop locally:
+
+``` bash
+git clone https://github.com/pytorch/fairseq
+cd fairseq
+pip install --editable ./
+
+# on MacOS:
+# CFLAGS="-stdlib=libc++" pip install --editable ./
+
+# to install the latest stable release (0.10.x)
+# pip install fairseq
+```
+
+* **For faster training** install NVIDIA's [apex](https://github.com/NVIDIA/apex) library:
+
+``` bash
+git clone https://github.com/NVIDIA/apex
+cd apex
+pip install -v --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" \
+  --global-option="--deprecated_fused_adam" --global-option="--xentropy" \
+  --global-option="--fast_multihead_attn" ./
+```
+
+* **For large datasets** install [PyArrow](https://arrow.apache.org/docs/python/install.html#using-pip): `pip install pyarrow`
+* If you use Docker make sure to increase the shared memory size either with `--ipc=host` or `--shm-size`
+ as command line options to `nvidia-docker run` .
+
+# Getting Started
+
+The [full documentation](https://fairseq.readthedocs.io/) contains instructions
+for getting started, training new models and extending fairseq with new model
+types and tasks.
+
+# Pre-trained models and examples
+
+We provide pre-trained models and pre-processed, binarized test sets for several tasks listed below,
+as well as example training and evaluation commands.
+
+* [Translation](examples/translation/README.md): convolutional and transformer models are available
+* [Language Modeling](examples/language_model/README.md): convolutional and transformer models are available
+
+We also have more detailed READMEs to reproduce results from specific papers:
+
+* [XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale (Babu et al., 2021)](examples/wav2vec/xlsr/README.md)
+* [Cross-lingual Retrieval for Iterative Self-Supervised Training (Tran et al., 2020)](examples/criss/README.md)
+* [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations (Baevski et al., 2020)](examples/wav2vec/README.md)
+* [Unsupervised Quality Estimation for Neural Machine Translation (Fomicheva et al., 2020)](examples/unsupervised_quality_estimation/README.md)
+* [Training with Quantization Noise for Extreme Model Compression ({Fan*, Stock*} et al., 2020)](examples/quant_noise/README.md)
+* [Neural Machine Translation with Byte-Level Subwords (Wang et al., 2020)](examples/byte_level_bpe/README.md)
+* [Multilingual Denoising Pre-training for Neural Machine Translation (Liu et at., 2020)](examples/mbart/README.md)
+* [Reducing Transformer Depth on Demand with Structured Dropout (Fan et al., 2019)](examples/layerdrop/README.md)
+* [Jointly Learning to Align and Translate with Transformer Models (Garg et al., 2019)](examples/joint_alignment_translation/README.md)
+* [Levenshtein Transformer (Gu et al., 2019)](examples/nonautoregressive_translation/README.md)
+* [Facebook FAIR's WMT19 News Translation Task Submission (Ng et al., 2019)](examples/wmt19/README.md)
+* [RoBERTa: A Robustly Optimized BERT Pretraining Approach (Liu et al., 2019)](examples/roberta/README.md)
+* [wav2vec: Unsupervised Pre-training for Speech Recognition (Schneider et al., 2019)](examples/wav2vec/README.md)
+* [Mixture Models for Diverse Machine Translation: Tricks of the Trade (Shen et al., 2019)](examples/translation_moe/README.md)
+* [Pay Less Attention with Lightweight and Dynamic Convolutions (Wu et al., 2019)](examples/pay_less_attention_paper/README.md)
+* [Understanding Back-Translation at Scale (Edunov et al., 2018)](examples/backtranslation/README.md)
+* [Classical Structured Prediction Losses for Sequence to Sequence Learning (Edunov et al., 2018)](https://github.com/pytorch/fairseq/tree/classic_seqlevel)
+* [Hierarchical Neural Story Generation (Fan et al., 2018)](examples/stories/README.md)
+* [Scaling Neural Machine Translation (Ott et al., 2018)](examples/scaling_nmt/README.md)
+* [Convolutional Sequence to Sequence Learning (Gehring et al., 2017)](examples/conv_seq2seq/README.md)
+* [Language Modeling with Gated Convolutional Networks (Dauphin et al., 2017)](examples/language_model/README.conv.md)
+
+# Join the fairseq community
+
+* Twitter: https://twitter.com/fairseq
+* Facebook page: https://www.facebook.com/groups/fairseq.users
+* Google group: https://groups.google.com/forum/#!forum/fairseq-users
+
+# License
+
+fairseq(-py) is MIT-licensed.
+The license applies to the pre-trained models as well.
+
+# Citation
+
+Please cite as:
+
+``` bibtex
+@inproceedings{ott2019fairseq,
+  title = {fairseq: A Fast, Extensible Toolkit for Sequence Modeling},
+  author = {Myle Ott and Sergey Edunov and Alexei Baevski and Angela Fan and Sam Gross and Nathan Ng and David Grangier and Michael Auli},
+  booktitle = {Proceedings of NAACL-HLT 2019: Demonstrations},
+  year = {2019},
+}
+```

PyTorch/NLP/new-Transformer/docs/Makefile

+# Minimal makefile for Sphinx documentation
+#
+
+# You can set these variables from the command line.
+SPHINXOPTS    =
+SPHINXBUILD   = python -msphinx
+SPHINXPROJ    = fairseq
+SOURCEDIR     = .
+BUILDDIR      = _build
+
+# Put it first so that "make" without argument is like "make help".
+help:
+	@$(SPHINXBUILD) -M help "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)
+
+.PHONY: help Makefile
+
+# Catch-all target: route all unknown targets to Sphinx using the new
+# "make mode" option.  $(O) is meant as a shortcut for $(SPHINXOPTS).
+%: Makefile
+	@$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)
\ No newline at end of file

PyTorch/NLP/new-Transformer/docs/_static/theme_overrides.css

+.wy-table-responsive table td kbd {
+    white-space: nowrap;
+}
+.wy-table-responsive table td {
+    white-space: normal !important;
+}
+.wy-table-responsive {
+    overflow: visible !important;
+}

PyTorch/NLP/new-Transformer/docs/command_line_tools.rst

+.. _Command-line Tools:
+
+Command-line Tools
+==================
+
+Fairseq provides several command-line tools for training and evaluating models:
+
+- :ref:`fairseq-preprocess`: Data pre-processing: build vocabularies and binarize training data
+- :ref:`fairseq-train`: Train a new model on one or multiple GPUs
+- :ref:`fairseq-generate`: Translate pre-processed data with a trained model
+- :ref:`fairseq-interactive`: Translate raw text with a trained model
+- :ref:`fairseq-score`: BLEU scoring of generated translations against reference translations
+- :ref:`fairseq-eval-lm`: Language model evaluation
+
+
+.. _fairseq-preprocess:
+
+fairseq-preprocess
+~~~~~~~~~~~~~~~~~~
+.. automodule:: fairseq_cli.preprocess
+
+    .. argparse::
+        :module: fairseq.options
+        :func: get_preprocessing_parser
+        :prog: fairseq-preprocess
+
+
+.. _fairseq-train:
+
+fairseq-train
+~~~~~~~~~~~~~
+.. automodule:: fairseq_cli.train
+
+    .. argparse::
+        :module: fairseq.options
+        :func: get_training_parser
+        :prog: fairseq-train
+
+
+.. _fairseq-generate:
+
+fairseq-generate
+~~~~~~~~~~~~~~~~
+.. automodule:: fairseq_cli.generate
+
+    .. argparse::
+        :module: fairseq.options
+        :func: get_generation_parser
+        :prog: fairseq-generate
+
+
+.. _fairseq-interactive:
+
+fairseq-interactive
+~~~~~~~~~~~~~~~~~~~
+.. automodule:: fairseq_cli.interactive
+
+    .. argparse::
+        :module: fairseq.options
+        :func: get_interactive_generation_parser
+        :prog: fairseq-interactive
+
+
+.. _fairseq-score:
+
+fairseq-score
+~~~~~~~~~~~~~
+.. automodule:: fairseq_cli.score
+
+    .. argparse::
+        :module: fairseq_cli.score
+        :func: get_parser
+        :prog: fairseq-score
+
+
+.. _fairseq-eval-lm:
+
+fairseq-eval-lm
+~~~~~~~~~~~~~~~
+.. automodule:: fairseq_cli.eval_lm
+
+    .. argparse::
+        :module: fairseq.options
+        :func: get_eval_lm_parser
+        :prog: fairseq-eval-lm

PyTorch/NLP/new-Transformer/docs/conf.py

+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+#
+# fairseq documentation build configuration file, created by
+# sphinx-quickstart on Fri Aug 17 21:45:30 2018.
+#
+# This file is execfile()d with the current directory set to its
+# containing dir.
+#
+# Note that not all possible configuration values are present in this
+# autogenerated file.
+#
+# All configuration values have a default; values that are commented out
+# serve to show the default.
+
+# If extensions (or modules to document with autodoc) are in another directory,
+# add these directories to sys.path here. If the directory is relative to the
+# documentation root, use os.path.abspath to make it absolute, like shown here.
+
+import os
+import sys
+from fairseq import __version__
+
+
+# source code directory, relative to this file, for sphinx-autobuild
+sys.path.insert(0, os.path.abspath(".."))
+
+source_suffix = [".rst"]
+
+# -- General configuration ------------------------------------------------
+
+# If your documentation needs a minimal Sphinx version, state it here.
+#
+# needs_sphinx = '1.0'
+
+# Add any Sphinx extension module names here, as strings. They can be
+# extensions coming with Sphinx (named 'sphinx.ext.*') or your custom
+# ones.
+extensions = [
+    "sphinx.ext.autodoc",
+    "sphinx.ext.intersphinx",
+    "sphinx.ext.viewcode",
+    "sphinx.ext.napoleon",
+    "sphinxarg.ext",
+]
+
+# Add any paths that contain templates here, relative to this directory.
+templates_path = ["_templates"]
+
+# The master toctree document.
+master_doc = "index"
+
+# General information about the project.
+project = "fairseq"
+copyright = "Facebook AI Research (FAIR)"
+author = "Facebook AI Research (FAIR)"
+
+github_doc_root = "https://github.com/pytorch/fairseq/tree/main/docs/"
+
+# The version info for the project you're documenting, acts as replacement for
+# |version| and |release|, also used in various other places throughout the
+# built documents.
+#
+# The short X.Y version.
+version = __version__
+# The full version, including alpha/beta/rc tags.
+release = __version__
+
+# The language for content autogenerated by Sphinx. Refer to documentation
+# for a list of supported languages.
+#
+# This is also used if you do content translation via gettext catalogs.
+# Usually you set "language" from the command line for these cases.
+language = None
+
+# List of patterns, relative to source directory, that match files and
+# directories to ignore when looking for source files.
+# This patterns also effect to html_static_path and html_extra_path
+exclude_patterns = ["_build", "Thumbs.db", ".DS_Store"]
+
+# The name of the Pygments (syntax highlighting) style to use.
+pygments_style = "sphinx"
+highlight_language = "python"
+
+# If true, `todo` and `todoList` produce output, else they produce nothing.
+todo_include_todos = False
+
+
+# -- Options for HTML output ----------------------------------------------
+
+# The theme to use for HTML and HTML Help pages.  See the documentation for
+# a list of builtin themes.
+#
+html_theme = "sphinx_rtd_theme"
+
+# Theme options are theme-specific and customize the look and feel of a theme
+# further.  For a list of options available for each theme, see the
+# documentation.
+#
+# html_theme_options = {}
+
+# Add any paths that contain custom static files (such as style sheets) here,
+# relative to this directory. They are copied after the builtin static files,
+# so a file named "default.css" will overwrite the builtin "default.css".
+html_static_path = ["_static"]
+
+html_context = {
+    "css_files": [
+        "_static/theme_overrides.css",  # override wide tables in RTD theme
+    ],
+}
+
+# Custom sidebar templates, must be a dictionary that maps document names
+# to template names.
+#
+# This is required for the alabaster theme
+# refs: http://alabaster.readthedocs.io/en/latest/installation.html#sidebars
+# html_sidebars = {
+#    '**': [
+#        'about.html',
+#        'navigation.html',
+#        'relations.html',  # needs 'show_related': True theme option to display
+#        'searchbox.html',
+#        'donate.html',
+#    ]
+# }
+
+
+# Example configuration for intersphinx: refer to the Python standard library.
+intersphinx_mapping = {
+    "numpy": ("http://docs.scipy.org/doc/numpy/", None),
+    "python": ("https://docs.python.org/", None),
+    "torch": ("https://pytorch.org/docs/master/", None),
+}

PyTorch/NLP/new-Transformer/docs/criterions.rst

+.. role:: hidden
+    :class: hidden-section
+
+.. _Criterions:
+
+Criterions
+==========
+
+Criterions compute the loss function given the model and batch, roughly::
+
+  loss = criterion(model, batch)
+
+.. automodule:: fairseq.criterions
+    :members:
+
+.. autoclass:: fairseq.criterions.FairseqCriterion
+    :members:
+    :undoc-members:
+
+.. autoclass:: fairseq.criterions.adaptive_loss.AdaptiveLoss
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.criterions.composite_loss.CompositeLoss
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.criterions.cross_entropy.CrossEntropyCriterion
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.criterions.label_smoothed_cross_entropy.LabelSmoothedCrossEntropyCriterion
+    :members:
+    :undoc-members:

PyTorch/NLP/new-Transformer/docs/data.rst

+.. role:: hidden
+    :class: hidden-section
+
+.. module:: fairseq.data
+
+Data Loading and Utilities
+==========================
+
+.. _datasets:
+
+Datasets
+--------
+
+**Datasets** define the data format and provide helpers for creating
+mini-batches.
+
+.. autoclass:: fairseq.data.FairseqDataset
+    :members:
+.. autoclass:: fairseq.data.LanguagePairDataset
+    :members:
+.. autoclass:: fairseq.data.MonolingualDataset
+    :members:
+
+**Helper Datasets**
+
+These datasets wrap other :class:`fairseq.data.FairseqDataset` instances and
+provide additional functionality:
+
+.. autoclass:: fairseq.data.BacktranslationDataset
+    :members:
+.. autoclass:: fairseq.data.ConcatDataset
+    :members:
+.. autoclass:: fairseq.data.ResamplingDataset
+    :members:
+.. autoclass:: fairseq.data.RoundRobinZipDatasets
+    :members:
+.. autoclass:: fairseq.data.TransformEosDataset
+    :members:
+
+
+Dictionary
+----------
+
+.. autoclass:: fairseq.data.Dictionary
+    :members:
+
+
+Iterators
+---------
+
+.. autoclass:: fairseq.data.CountingIterator
+    :members:
+.. autoclass:: fairseq.data.EpochBatchIterator
+    :members:
+.. autoclass:: fairseq.data.GroupedIterator
+    :members:
+.. autoclass:: fairseq.data.ShardedIterator
+    :members:

PyTorch/NLP/new-Transformer/docs/docutils.conf

+[writers]
+option-limit=0

PyTorch/NLP/new-Transformer/docs/getting_started.rst

+Evaluating Pre-trained Models
+=============================
+
+First, download a pre-trained model along with its vocabularies:
+
+.. code-block:: console
+
+    > curl https://dl.fbaipublicfiles.com/fairseq/models/wmt14.v2.en-fr.fconv-py.tar.bz2 | tar xvjf -
+
+This model uses a `Byte Pair Encoding (BPE)
+vocabulary <https://arxiv.org/abs/1508.07909>`__, so we'll have to apply
+the encoding to the source text before it can be translated. This can be
+done with the
+`apply\_bpe.py <https://github.com/rsennrich/subword-nmt/blob/master/subword_nmt/apply_bpe.py>`__
+script using the ``wmt14.en-fr.fconv-cuda/bpecodes`` file. ``@@`` is
+used as a continuation marker and the original text can be easily
+recovered with e.g. ``sed s/@@ //g`` or by passing the ``--remove-bpe``
+flag to :ref:`fairseq-generate`. Prior to BPE, input text needs to be tokenized
+using ``tokenizer.perl`` from
+`mosesdecoder <https://github.com/moses-smt/mosesdecoder>`__.
+
+Let's use :ref:`fairseq-interactive` to generate translations interactively.
+Here, we use a beam size of 5 and preprocess the input with the Moses
+tokenizer and the given Byte-Pair Encoding vocabulary. It will automatically
+remove the BPE continuation markers and detokenize the output.
+
+.. code-block:: console
+
+    > MODEL_DIR=wmt14.en-fr.fconv-py
+    > fairseq-interactive \
+        --path $MODEL_DIR/model.pt $MODEL_DIR \
+        --beam 5 --source-lang en --target-lang fr \
+        --tokenizer moses \
+        --bpe subword_nmt --bpe-codes $MODEL_DIR/bpecodes
+    | loading model(s) from wmt14.en-fr.fconv-py/model.pt
+    | [en] dictionary: 44206 types
+    | [fr] dictionary: 44463 types
+    | Type the input sentence and press return:
+    Why is it rare to discover new marine mammal species?
+    S-0     Why is it rare to discover new marine mam@@ mal species ?
+    H-0     -0.0643349438905716     Pourquoi est-il rare de découvrir de nouvelles espèces de mammifères marins?
+    P-0     -0.0763 -0.1849 -0.0956 -0.0946 -0.0735 -0.1150 -0.1301 -0.0042 -0.0321 -0.0171 -0.0052 -0.0062 -0.0015
+
+This generation script produces three types of outputs: a line prefixed
+with *O* is a copy of the original source sentence; *H* is the
+hypothesis along with an average log-likelihood; and *P* is the
+positional score per token position, including the
+end-of-sentence marker which is omitted from the text.
+
+Other types of output lines you might see are *D*, the detokenized hypothesis,
+*T*, the reference target, *A*, alignment info, *E* the history of generation steps.
+
+See the `README <https://github.com/pytorch/fairseq#pre-trained-models>`__ for a
+full list of pre-trained models available.
+
+Training a New Model
+====================
+
+The following tutorial is for machine translation. For an example of how
+to use Fairseq for other tasks, such as :ref:`language modeling`, please see the
+``examples/`` directory.
+
+Data Pre-processing
+-------------------
+
+Fairseq contains example pre-processing scripts for several translation
+datasets: IWSLT 2014 (German-English), WMT 2014 (English-French) and WMT
+2014 (English-German). To pre-process and binarize the IWSLT dataset:
+
+.. code-block:: console
+
+    > cd examples/translation/
+    > bash prepare-iwslt14.sh
+    > cd ../..
+    > TEXT=examples/translation/iwslt14.tokenized.de-en
+    > fairseq-preprocess --source-lang de --target-lang en \
+        --trainpref $TEXT/train --validpref $TEXT/valid --testpref $TEXT/test \
+        --destdir data-bin/iwslt14.tokenized.de-en
+
+This will write binarized data that can be used for model training to
+``data-bin/iwslt14.tokenized.de-en``.
+
+Training
+--------
+
+Use :ref:`fairseq-train` to train a new model. Here a few example settings that work
+well for the IWSLT 2014 dataset:
+
+.. code-block:: console
+
+    > mkdir -p checkpoints/fconv
+    > CUDA_VISIBLE_DEVICES=0 fairseq-train data-bin/iwslt14.tokenized.de-en \
+        --optimizer nag --lr 0.25 --clip-norm 0.1 --dropout 0.2 --max-tokens 4000 \
+        --arch fconv_iwslt_de_en --save-dir checkpoints/fconv
+
+By default, :ref:`fairseq-train` will use all available GPUs on your machine. Use the
+``CUDA_VISIBLE_DEVICES`` environment variable to select specific GPUs and/or to
+change the number of GPU devices that will be used.
+
+Also note that the batch size is specified in terms of the maximum
+number of tokens per batch (``--max-tokens``). You may need to use a
+smaller value depending on the available GPU memory on your system.
+
+Generation
+----------
+
+Once your model is trained, you can generate translations using
+:ref:`fairseq-generate` **(for binarized data)** or
+:ref:`fairseq-interactive` **(for raw text)**:
+
+.. code-block:: console
+
+    > fairseq-generate data-bin/iwslt14.tokenized.de-en \
+        --path checkpoints/fconv/checkpoint_best.pt \
+        --batch-size 128 --beam 5
+    | [de] dictionary: 35475 types
+    | [en] dictionary: 24739 types
+    | data-bin/iwslt14.tokenized.de-en test 6750 examples
+    | model fconv
+    | loaded checkpoint trainings/fconv/checkpoint_best.pt
+    S-721   danke .
+    T-721   thank you .
+    ...
+
+To generate translations with only a CPU, use the ``--cpu`` flag. BPE
+continuation markers can be removed with the ``--remove-bpe`` flag.
+
+Advanced Training Options
+=========================
+
+Large mini-batch training with delayed updates
+----------------------------------------------
+
+The ``--update-freq`` option can be used to accumulate gradients from
+multiple mini-batches and delay updating, creating a larger effective
+batch size. Delayed updates can also improve training speed by reducing
+inter-GPU communication costs and by saving idle time caused by variance
+in workload across GPUs. See `Ott et al.
+(2018) <https://arxiv.org/abs/1806.00187>`__ for more details.
+
+To train on a single GPU with an effective batch size that is equivalent
+to training on 8 GPUs:
+
+.. code-block:: console
+
+    > CUDA_VISIBLE_DEVICES=0 fairseq-train --update-freq 8 (...)
+
+Training with half precision floating point (FP16)
+--------------------------------------------------
+
+.. note::
+
+    FP16 training requires a Volta GPU and CUDA 9.1 or greater
+
+Recent GPUs enable efficient half precision floating point computation,
+e.g., using `Nvidia Tensor Cores
+<https://docs.nvidia.com/deeplearning/sdk/mixed-precision-training/index.html>`__.
+Fairseq supports FP16 training with the ``--fp16`` flag:
+
+.. code-block:: console
+
+    > fairseq-train --fp16 (...)
+
+Distributed training
+--------------------
+
+Distributed training in fairseq is implemented on top of ``torch.distributed``.
+The easiest way to launch jobs is with the `torch.distributed.launch
+<https://pytorch.org/docs/stable/distributed.html#launch-utility>`__ tool.
+
+For example, to train a large English-German Transformer model on 2 nodes each
+with 8 GPUs (in total 16 GPUs), run the following command on each node,
+replacing ``node_rank=0`` with ``node_rank=1`` on the second node and making
+sure to update ``--master_addr`` to the IP address of the first node:
+
+.. code-block:: console
+
+    > python -m torch.distributed.launch --nproc_per_node=8 \
+        --nnodes=2 --node_rank=0 --master_addr="192.168.1.1" \
+        --master_port=12345 \
+        $(which fairseq-train) data-bin/wmt16_en_de_bpe32k \
+        --arch transformer_vaswani_wmt_en_de_big --share-all-embeddings \
+        --optimizer adam --adam-betas '(0.9, 0.98)' --clip-norm 0.0 \
+        --lr-scheduler inverse_sqrt --warmup-init-lr 1e-07 --warmup-updates 4000 \
+        --lr 0.0005 \
+        --dropout 0.3 --weight-decay 0.0 --criterion label_smoothed_cross_entropy --label-smoothing 0.1 \
+        --max-tokens 3584 \
+        --max-epoch 70 \
+        --fp16
+
+On SLURM clusters, fairseq will automatically detect the number of nodes and
+GPUs, but a port number must be provided:
+
+.. code-block:: console
+
+    > salloc --gpus=16 --nodes 2 (...)
+    > srun fairseq-train --distributed-port 12345 (...).
+
+Sharding very large datasets
+----------------------------
+
+It can be challenging to train over very large datasets, particularly if your
+machine does not have much system RAM. Most tasks in fairseq support training
+over "sharded" datasets, in which the original dataset has been preprocessed
+into non-overlapping chunks (or "shards").
+
+For example, instead of preprocessing all your data into a single "data-bin"
+directory, you can split the data and create "data-bin1", "data-bin2", etc.
+Then you can adapt your training command like so:
+
+.. code-block:: console
+
+    > fairseq-train data-bin1:data-bin2:data-bin3 (...)
+
+Training will now iterate over each shard, one by one, with each shard
+corresponding to an "epoch", thus reducing system memory usage.

PyTorch/NLP/new-Transformer/docs/hydra_integration.md

+## Hydra
+
+[Hydra](https://github.com/facebookresearch/hydra) is an open-source Python
+framework that simplifies the development of research and other complex
+applications. The key feature is the ability to dynamically create a
+hierarchical configuration by composition and override it through config files
+and the command line. The name Hydra comes from its ability to run multiple
+similar jobs - much like a Hydra with multiple heads.
+
+## Motivation
+
+Until recently, all components in fairseq were configured through a shared
+`args` namespace that was created at application startup. Components declared
+their own `add_args` method to update the argparse parser, hoping that the names
+would not clash with arguments from other components. While this model works for
+smaller applications, as fairseq grew and became integrated into other
+applications, this became problematic. In order to determine how to configure
+each component, one needed to a) examine what args were added by this component,
+and b) read the code to figure out what shared arguments it is using that were
+added in other places. Reproducing models involved sharing commands that often
+contained dozens of command line switches.
+
+The model described above is still supported by fairseq for backward
+compatibility, but will be deprecated some time in the future.
+
+New components in fairseq should now create a dataclass that encapsulates all
+parameters required to configure this component. The dataclass is registered
+along with the component, and fairseq takes care of constructing and providing
+this configuration object to the component's constructor. Note that sharing
+parameters can optionally still work, but one has to explicitly point to the
+"source of truth" (see inheritance example below). These changes make components
+in fairseq more independent and re-usable by other applications: all that is
+needed to create a component is to initialize its dataclass and overwrite some
+of the defaults.
+
+While configuring fairseq through command line (using either the legacy argparse
+based or the new Hydra based entry points) is still fully supported, you can now
+take advantage of configuring fairseq completely or piece-by-piece through
+hierarchical YAML configuration files. These files can also be shipped as
+examples that others can use to run an identically configured job.
+
+Additionally, Hydra has a rich and growing [library of
+plugins](https://github.com/facebookresearch/hydra/tree/master/plugins) that
+provide functionality such as hyperparameter sweeping (including using bayesian
+optimization through the [Ax](https://github.com/facebook/Ax) library), job
+launching across various platforms, and more.
+
+## Creating or migrating components
+
+In general, each new (or updated) component should provide a companion
+[dataclass](https://www.python.org/dev/peps/pep-0557/). These dataclass are
+typically located in the same file as the component and are passed as arguments
+to the `register_*()` functions. Top-level configs that should be present in
+every fairseq application are placed in the
+[global](fairseq/dataclass/configs.py) config file and added to the
+`FairseqConfig` object.
+
+Each dataclass is a plain-old-data object, similar to a `NamedTuple`. These
+classes are decorated with a `@dataclass` decorator, and typically inherit from
+`FairseqDataclass` (which adds some functionality for backward compatibility).
+Each field must have a type, and generally has metadata (such as a help string)
+and a default value. Only primitive types or other config objects are allowed as
+data types for each field.
+
+#### Example:
+
+```python
+from dataclasses import dataclass, field
+from fairseq.dataclass import FairseqDataclass
+
+@dataclass
+class InteractiveConfig(FairseqDataclass):
+    buffer_size: int = field(
+        default=0,
+        metadata={
+            "help": "read this many sentences into a buffer before processing them"
+        },
+    )
+    input: str = field(
+        default="-",
+        metadata={"help": "file to read from; use - for stdin"},
+    )
+```
+
+### Inherting values
+
+Some components require sharing a value. For example, a learning rate scheduler
+and an optimizer may both need to know the initial learning rate value. One can
+declare a field that, by default, will inherit its value from another config
+node in the same hierarchy:
+
+```python
+@dataclass
+FairseqAdamConfig(FairseqDataclass):
+    ...
+    lr: List[float] = II("optimization.lr")
+    ...
+```
+
+`II("optimization.lr")` is syntactic sugar for `"${optimization.lr}"`, which is
+the value one can use in a YAML config file or through command line to achieve
+the same effect. Note that this assumes that there is an "optimization" config
+object in the root config and it has a field called "lr".
+
+### Tasks and Models
+
+Creating Tasks and Models works same as before, except that legacy
+implementations now inherit from `LegacyFairseq*` base classes, while new
+components inherit from `FairseqTask` and `FairseqModel` and provide a dataclass
+to the `register_*()` functions.
+
+#### Task example:
+
+```python
+@dataclass
+class LanguageModelingConfig(FairseqDataclass):
+    data: Optional[str] = field(
+        default=None, metadata={"help": "path to data directory"}
+    )
+    ...
+
+@register_task("language_modeling", dataclass=LanguageModelingConfig)
+class LanguageModelingTask(FairseqTask):
+    ...
+    @classmethod
+    def setup_task(cls, cfg: LanguageModelingConfig):
+        ...
+```
+
+#### Model example:
+
+```python
+@dataclass
+class TransformerLanguageModelConfig(FairseqDataclass):
+    activation_fn: ChoiceEnum(utils.get_available_activation_fns()) = field(
+        default="relu", metadata={"help": "activation function to use"}
+    )
+    dropout: float = field(default=0.1, metadata={"help": "dropout probability"})
+    ...
+
+@register_model("transformer_lm", dataclass=TransformerLanguageModelConfig)
+class TransformerLanguageModel(FairseqLanguageModel):
+    ...
+    @classmethod
+    def build_model(cls, cfg: TransformerLanguageModelConfig, task: FairseqTask):
+        ...
+```
+
+### Other components
+
+Other components work as before, but they now take their configuration dataclass
+as the only constructor argument:
+
+```python
+@dataclass
+class MosesTokenizerConfig(FairseqDataclass):
+    source_lang: str = field(default="en", metadata={"help": "source language"})
+    ...
+
+@register_tokenizer("moses", dataclass=MosesTokenizerConfig)
+class MosesTokenizer(object):
+    def __init__(self, cfg: MosesTokenizerConfig):
+        ...
+```
+
+Note that if you are adding a new registry for a new set of components, you need
+to add it to the `FairseqConfig` object in `fairseq/dataclass/configs.py`:
+
+```python
+@dataclass
+class FairseqConfig(object):
+    ...
+    my_new_registry: Any = None
+```
+
+## Training with `fairseq-hydra-train`
+
+To fully take advantage of configuration flexibility offered by Hydra, you may
+want to train new models using the `fairseq-hydra-train` entry point. Legacy CLI
+tools such as `fairseq-train` will remain supported for the foreseeable future
+but will be deprecated eventually.
+
+On startup, Hydra will create a configuration object that contains a hierarchy
+of all the necessary dataclasses populated with their default values in the
+code. The default values are overwritten by values found in YAML files in
+`fairseq/config` directory (which currently sets minimal defaults) and then
+further overwritten by values provided through command line arguments.
+
+Some of the most common use cases are shown below:
+
+### 1. Override default values through command line:
+
+```shell script
+$ fairseq-hydra-train \
+    distributed_training.distributed_world_size=1 \
+    dataset.batch_size=2 \
+    task.data=data-bin \
+    model=transformer_lm/transformer_lm_gpt \
+    task=language_modeling \
+    optimization.max_update=5000
+```
+
+Note that along with explicitly providing values for parameters such as
+`dataset.batch_size`, this also tells Hydra to overlay configuration found in
+`fairseq/config/model/transformer_lm/transformer_lm_gpt.yaml` over the default
+values in the dataclass. If you want to train a model without specifying a
+particular architecture you can simply specify `model=transformer_lm`. This only
+works for migrated tasks and models.
+
+### 2. Replace bundled configs with an external config:
+
+```shell script
+$ fairseq-hydra-train \
+    --config-dir /path/to/external/configs \
+    --config-name wiki103
+```
+
+where `/path/to/external/configs/wiki103.yaml` contains:
+
+```yaml
+# @package _group_
+
+model:
+  _name: transformer_lm
+distributed_training:
+  distributed_world_size: 1
+dataset:
+  batch_size: 2
+task:
+  _name: language_modeling
+  data: /path/to/data
+  add_bos_token: false
+  max_target_positions: 1024
+optimization:
+  max_update: 50000
+  lr: [ 0.25 ]
+criterion: cross_entropy
+optimizer: adam
+lr_scheduler:
+  _name: cosine
+```
+
+Note that here bundled configs from `fairseq/config` directory are not used,
+however the defaults from each dataclass will still be used (unless overwritten
+by your external config).
+
+Additionally you can choose to break up your configs by creating a directory
+structure in the same location as your main config file, with the names of the
+top-level fields (such as "model", "dataset", etc), and placing config files
+with meaningful names that would populate that specific section of your
+top-level config file (for example, you might have
+`model/small_transformer_lm.yaml`, `model/big_transformer_lm.yaml`, etc). You
+can then specify the correct configuration via command line, defaults in the
+main config, or even launch all of them as a sweep (see Hydra documentation on
+how to do this).
+
+### 3. Add an external config directory to Hydra search path:
+
+This allows combining default configuration (including using any bundled config
+files), while specifying your own config files for some parts of the
+configuration.
+
+```shell script
+$ fairseq-hydra-train \
+    distributed_training.distributed_world_size=1 \
+    dataset.batch_size=2 \
+    task.data=/path/to/data/ \
+    model=transformer_lm/2_layers \
+    task=language_modeling \
+    optimization.max_update=5000 \
+    --config-dir /path/to/external/configs
+```
+
+where `/path/to/external/configs` has the following structure:
+```
+.
++-- model
+|   +-- transformer_lm
+|   |   +-- 2_layers.yaml
+```
+
+and `2_layers.yaml` contains a copy of `transformer_lm_gpt.yaml` but with
+`decoder_layers` set to 2. You can add other configs to configure other
+components as well.

PyTorch/NLP/new-Transformer/docs/index.rst

+.. fairseq documentation master file, created by
+   sphinx-quickstart on Fri Aug 17 21:45:30 2018.
+   You can adapt this file completely to your liking, but it should at least
+   contain the root `toctree` directive.
+
+:github_url: https://github.com/pytorch/fairseq
+
+
+fairseq documentation
+=====================
+
+Fairseq is a sequence modeling toolkit written in `PyTorch
+<http://pytorch.org/>`_ that allows researchers and developers to
+train custom models for translation, summarization, language modeling and other
+text generation tasks.
+
+.. toctree::
+    :maxdepth: 1
+    :caption: Getting Started
+
+    getting_started
+    command_line_tools
+
+.. toctree::
+    :maxdepth: 1
+    :caption: Extending Fairseq
+
+    overview
+    tutorial_simple_lstm
+    tutorial_classifying_names
+
+.. toctree::
+    :maxdepth: 2
+    :caption: Library Reference
+
+    tasks
+    models
+    criterions
+    optim
+    lr_scheduler
+    data
+    modules
+
+
+Indices and tables
+==================
+
+* :ref:`genindex`
+* :ref:`search`

PyTorch/NLP/new-Transformer/docs/lr_scheduler.rst

+.. role:: hidden
+    :class: hidden-section
+
+.. _Learning Rate Schedulers:
+
+Learning Rate Schedulers
+========================
+
+Learning Rate Schedulers update the learning rate over the course of training.
+Learning rates can be updated after each update via :func:`step_update` or at
+epoch boundaries via :func:`step`.
+
+.. automodule:: fairseq.optim.lr_scheduler
+    :members:
+
+.. autoclass:: fairseq.optim.lr_scheduler.FairseqLRScheduler
+    :members:
+    :undoc-members:
+
+.. autoclass:: fairseq.optim.lr_scheduler.cosine_lr_scheduler.CosineSchedule
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.optim.lr_scheduler.fixed_schedule.FixedSchedule
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.optim.lr_scheduler.inverse_square_root_schedule.InverseSquareRootSchedule
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.optim.lr_scheduler.reduce_lr_on_plateau.ReduceLROnPlateau
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.optim.lr_scheduler.triangular_lr_scheduler.TriangularSchedule
+    :members:
+    :undoc-members:

PyTorch/NLP/new-Transformer/docs/make.bat

+@ECHO OFF
+
+pushd %~dp0
+
+REM Command file for Sphinx documentation
+
+if "%SPHINXBUILD%" == "" (
+	set SPHINXBUILD=python -msphinx
+)
+set SOURCEDIR=.
+set BUILDDIR=_build
+set SPHINXPROJ=fairseq
+
+if "%1" == "" goto help
+
+%SPHINXBUILD% >NUL 2>NUL
+if errorlevel 9009 (
+	echo.
+	echo.The Sphinx module was not found. Make sure you have Sphinx installed,
+	echo.then set the SPHINXBUILD environment variable to point to the full
+	echo.path of the 'sphinx-build' executable. Alternatively you may add the
+	echo.Sphinx directory to PATH.
+	echo.
+	echo.If you don't have Sphinx installed, grab it from
+	echo.http://sphinx-doc.org/
+	exit /b 1
+)
+
+%SPHINXBUILD% -M %1 %SOURCEDIR% %BUILDDIR% %SPHINXOPTS%
+goto end
+
+:help
+%SPHINXBUILD% -M help %SOURCEDIR% %BUILDDIR% %SPHINXOPTS%
+
+:end
+popd

PyTorch/NLP/new-Transformer/docs/models.rst

+.. role:: hidden
+    :class: hidden-section
+
+.. module:: fairseq.models
+
+.. _Models:
+
+Models
+======
+
+A Model defines the neural network's ``forward()`` method and encapsulates all
+of the learnable parameters in the network. Each model also provides a set of
+named *architectures* that define the precise network configuration (e.g.,
+embedding dimension, number of layers, etc.).
+
+Both the model type and architecture are selected via the ``--arch``
+command-line argument. Once selected, a model may expose additional command-line
+arguments for further configuration.
+
+.. note::
+
+    All fairseq Models extend :class:`BaseFairseqModel`, which in turn extends
+    :class:`torch.nn.Module`. Thus any fairseq Model can be used as a
+    stand-alone Module in other PyTorch code.
+
+
+Convolutional Neural Networks (CNN)
+-----------------------------------
+
+.. module:: fairseq.models.fconv
+.. autoclass:: fairseq.models.fconv.FConvModel
+    :members:
+.. autoclass:: fairseq.models.fconv.FConvEncoder
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.models.fconv.FConvDecoder
+    :members:
+
+
+Long Short-Term Memory (LSTM) networks
+--------------------------------------
+
+.. module:: fairseq.models.lstm
+.. autoclass:: fairseq.models.lstm.LSTMModel
+    :members:
+.. autoclass:: fairseq.models.lstm.LSTMEncoder
+    :members:
+.. autoclass:: fairseq.models.lstm.LSTMDecoder
+    :members:
+
+
+Transformer (self-attention) networks
+-------------------------------------
+
+.. module:: fairseq.models.transformer
+.. autoclass:: fairseq.models.transformer.TransformerModel
+    :members:
+.. autoclass:: fairseq.models.transformer.TransformerEncoder
+    :members:
+.. autoclass:: fairseq.models.transformer.TransformerEncoderLayer
+    :members:
+.. autoclass:: fairseq.models.transformer.TransformerDecoder
+    :members:
+.. autoclass:: fairseq.models.transformer.TransformerDecoderLayer
+    :members:
+
+
+Adding new models
+-----------------
+
+.. currentmodule:: fairseq.models
+.. autofunction:: fairseq.models.register_model
+.. autofunction:: fairseq.models.register_model_architecture
+.. autoclass:: fairseq.models.BaseFairseqModel
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.models.FairseqEncoderDecoderModel
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.models.FairseqEncoderModel
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.models.FairseqLanguageModel
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.models.FairseqMultiModel
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.models.FairseqEncoder
+    :members:
+.. autoclass:: fairseq.models.CompositeEncoder
+    :members:
+.. autoclass:: fairseq.models.FairseqDecoder
+    :members:
+
+
+.. _Incremental decoding:
+
+Incremental decoding
+--------------------
+
+.. autoclass:: fairseq.models.FairseqIncrementalDecoder
+    :members:
+    :undoc-members:

PyTorch/NLP/new-Transformer/docs/modules.rst

+Modules
+=======
+
+Fairseq provides several stand-alone :class:`torch.nn.Module` classes that may
+be helpful when implementing a new :class:`~fairseq.models.BaseFairseqModel`.
+
+.. automodule:: fairseq.modules
+    :members:
+    :undoc-members:

PyTorch/NLP/new-Transformer/docs/optim.rst

+.. role:: hidden
+    :class: hidden-section
+
+.. _optimizers:
+
+Optimizers
+==========
+
+Optimizers update the Model parameters based on the gradients.
+
+.. automodule:: fairseq.optim
+    :members:
+
+.. autoclass:: fairseq.optim.FairseqOptimizer
+    :members:
+    :undoc-members:
+
+.. autoclass:: fairseq.optim.adadelta.Adadelta
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.optim.adagrad.Adagrad
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.optim.adafactor.FairseqAdafactor
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.optim.adam.FairseqAdam
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.optim.fp16_optimizer.FP16Optimizer
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.optim.nag.FairseqNAG
+    :members:
+    :undoc-members:
+.. autoclass:: fairseq.optim.sgd.SGD
+    :members:
+    :undoc-members: