DeepLearningExamples/PyTorch/Translation/GNMT

GNMT v2 For PyTorch

This repository provides a script and recipe to train the GNMT v2 model to achieve state of the art accuracy, and is tested and maintained by NVIDIA.

Table Of Contents

• Model overview
  • Model architecture
  • Default configuration
  • Feature support matrix
    • Features
  • Mixed precision training
    • Enabling mixed precision
• Setup
  • Requirements
• Quick Start Guide
• Advanced
  • Scripts and sample code
  • Parameters
  • Command-line options
  • Getting the data
    • Dataset guidelines
  • Training process
  • Inference process
• Performance
  • Benchmarking
    • Training performance benchmark
    • Inference performance benchmark
  • Results
    • Training accuracy results
      • Training accuracy: NVIDIA DGX-1 (8x V100 16G)
      • Training accuracy: NVIDIA DGX-2 (16x V100 32G)
      • Training stability test
    • Training throughput results
      • Training throughput: NVIDIA DGX-1 (8x V100 16G)
      • Training throughput: NVIDIA DGX-2 (16x V100 32G)
    • Inference accuracy results
      • Inference accuracy: NVIDIA Tesla V100 16G
    • Inference throughput results
      • Inference throughput: NVIDIA Tesla V100 16G
    • Inference latency results
      • Inference latency: NVIDIA Tesla V100 16G
• Release notes
  • Changelog
  • Known issues

Model overview

The GNMT v2 model is similar to the one discussed in the Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation paper.

The most important difference between the two models is in the attention mechanism. In our model, the output from the first LSTM layer of the decoder goes into the attention module, then the re-weighted context is concatenated with inputs to all subsequent LSTM layers in the decoder at the current time step.

The same attention mechanism is also implemented in the default GNMT-like models from TensorFlow Neural Machine Translation Tutorial and NVIDIA OpenSeq2Seq Toolkit.

Model architecture

Default configuration

The following features were implemented in this model:

• general:
  • encoder and decoder are using shared embeddings
  • data-parallel multi-GPU training
  • dynamic loss scaling with backoff for Tensor Cores (mixed precision) training
  • trained with label smoothing loss (smoothing factor 0.1)
• encoder:
  • 4-layer LSTM, hidden size 1024, first layer is bidirectional, the rest are unidirectional
  • with residual connections starting from 3rd layer
  • uses standard PyTorch nn.LSTM layer
  • dropout is applied on input to all LSTM layers, probability of dropout is set to 0.2
  • hidden state of LSTM layers is initialized with zeros
  • weights and bias of LSTM layers is initialized with uniform(-0.1,0.1) distribution
• decoder:
  • 4-layer unidirectional LSTM with hidden size 1024 and fully-connected classifier
  • with residual connections starting from 3rd layer
  • uses standard PyTorch nn.LSTM layer
  • dropout is applied on input to all LSTM layers, probability of dropout is set to 0.2
  • hidden state of LSTM layers is initialized with zeros
  • weights and bias of LSTM layers is initialized with uniform(-0.1,0.1) distribution
  • weights and bias of fully-connected classifier is initialized with uniform(-0.1,0.1) distribution
• attention:
  • normalized Bahdanau attention
  • output from first LSTM layer of decoder goes into attention, then re-weighted context is concatenated with the input to all subsequent LSTM layers of the decoder at the current timestep
  • linear transform of keys and queries is initialized with uniform(-0.1, 0.1), normalization scalar is initialized with 1.0/sqrt(1024), normalization bias is initialized with zero
• inference:
  • beam search with default beam size of 5
  • with coverage penalty and length normalization, coverage penalty factor is set to 0.1, length normalization factor is set to 0.6 and length normalization constant is set to 5.0
  • de-tokenized BLEU computed by SacreBLEU
  • motivation for choosing SacreBLEU

When comparing the BLEU score, there are various tokenization approaches and BLEU calculation methodologies; therefore, ensure you align similar metrics.

Code from this repository can be used to train a larger, 8-layer GNMT v2 model. Our experiments show that a 4-layer model is significantly faster to train and yields comparable accuracy on the public WMT16 English-German dataset. The number of LSTM layers is controlled by the --num_layers parameter in the train.py training script.

Feature support matrix

The following features are supported by this model.

Feature	GNMT v2
Apex AMP	Yes
Apex DistributedDataParallel	Yes

Features

Apex AMP - a tool that enables Tensor Core-accelerated training. Refer to the Enabling mixed precision section for more details.

Apex DistributedDataParallel - a module wrapper that enables easy multiprocess distributed data parallel training, similar to torch.nn.parallel.DistributedDataParallel. DistributedDataParallel is optimized for use with NCCL. It achieves high performance by overlapping communication with computation during backward() and bucketing smaller gradient transfers to reduce the total number of transfers required.

Mixed precision training

Mixed precision is the combined use of different numerical precisions in a computational method. Mixed precision training offers significant computational speedup by performing operations in half-precision format, while storing minimal information in single-precision to retain as much information as possible in critical parts of the network. Since the introduction of Tensor Cores in the Volta and Turing architectures, significant training speedups are experienced by switching to mixed precision -- up to 3x overall speedup on the most arithmetically intense model architectures. Using mixed precision training previously required two steps:

1. Porting the model to use the FP16 data type where appropriate.
2. Manually adding loss scaling to preserve small gradient values.

The ability to train deep learning networks with lower precision was introduced in the Pascal architecture and first supported in CUDA 8 in the NVIDIA Deep Learning SDK.

For information about:

• How to train using mixed precision, see the Mixed Precision Training paper and Training With Mixed Precision documentation.
• Techniques used for mixed precision training, see the Mixed-Precision Training of Deep Neural Networks blog.
• APEX tools for mixed precision training, see the NVIDIA Apex: Tools for Easy Mixed-Precision Training in PyTorch .

Enabling mixed precision

By default, the train.py training script will launch mixed precision training with Tensor Cores. You can change this behavior and execute the training in single precision by setting the --math fp32 flag for the train.py training script.

Mixed precision is enabled in PyTorch by using the Automatic Mixed Precision (AMP), library from APEX that casts variables to half-precision upon retrieval, while storing variables in single-precision format. Furthermore, to preserve small gradient magnitudes in backpropagation, a loss scaling step must be included when applying gradients. In PyTorch, loss scaling can be easily applied by using scale_loss() method provided by AMP. The scaling value to be used can be dynamic or fixed.

For an in-depth walk through on AMP, check out sample usage here. APEX is a PyTorch extension that contains utility libraries, such as AMP, which require minimal network code changes to leverage tensor cores performance.

The following steps were needed to enable mixed precision training in GNMT:

• Import AMP from APEX (file: seq2seq/train/trainer.py):

from apex import amp

• Initialize AMP and wrap the model and the optimizer (file: seq2seq/train/trainer.py, class: Seq2SeqTrainer):

self.model, self.optimizer = amp.initialize(
    self.model,
    self.optimizer,
    cast_model_outputs=torch.float16,
    keep_batchnorm_fp32=False,
    opt_level='O2')

• Apply scale_loss context manager (file: seq2seq/train/fp_optimizers.py, class: AMPOptimizer):

with amp.scale_loss(loss, optimizer) as scaled_loss:
    scaled_loss.backward()

• Apply gradient clipping on single precision master weights (file: seq2seq/train/fp_optimizers.py, class: AMPOptimizer):

if self.grad_clip != float('inf'):
    clip_grad_norm_(amp.master_params(optimizer), self.grad_clip)

Setup

The following section lists the requirements in order to start training the GNMT v2 model.

Requirements

This repository contains Dockerfile which extends the PyTorch NGC container and encapsulates some dependencies. Aside from these dependencies, ensure you have the following components:

• NVIDIA Docker
• PyTorch 19.05-py3 NGC container
• NVIDIA Volta or Turing based GPU

For more information about how to get started with NGC containers, see the following sections from the NVIDIA GPU Cloud Documentation and the Deep Learning DGX Documentation:

• Getting Started Using NVIDIA GPU Cloud,
• Accessing And Pulling From The NGC container registry,
• Running PyTorch.

For those unable to use the Pytorch NGC container, to set up the required environment or create your own container, see the versioned NVIDIA Container Support Matrix.

Quick Start Guide

To train your model using mixed precision with Tensor Cores or using FP32, perform the following steps using the default parameters of the GNMT v2 model on the WMT16 English German dataset. For the specifics concerning training and inference, see the Advanced section.

1. Clone the repository.

git clone https://github.com/NVIDIA/DeepLearningExamples
cd DeepLearningExamples/PyTorch/Translation/GNMT

2. Build the GNMT v2 Docker container.

bash scripts/docker/build.sh

3. Start an interactive session in the container to run training/inference.

bash scripts/docker/interactive.sh

4. Download and preprocess the dataset.

Data will be downloaded to the data directory (on the host). The data directory is mounted to the /workspace/gnmt/data location in the Docker container.

bash scripts/wmt16_en_de.sh

5. Start training.

By default, the train.py training script will use all available GPUs. The training script saves only one checkpoint with the lowest value of the loss function on the validation dataset. All results and logs are saved to the results directory (on the host) or to the /workspace/gnmt/results directory (in the container). By default, the train.py script will launch mixed precision training with Tensor Cores. You can change this behavior by setting the --math fp32 flag for the train.py training script.

To launch mixed precision training on 1, 4 or 8 GPUs, run:

python3 -m launch train.py --seed 2 --train-global-batch-size 1024

To launch mixed precision training on 16 GPUs, run:

python3 -m launch train.py --seed 2 --train-global-batch-size 2048

By default, the training script will launch training with batch size 128 per GPU. If --train-global-batch-size is specified and larger than 128 times the number of GPUs available for the training then the training script will accumulate gradients over consecutive iterations and then perform the weight update. For example, 1 GPU training with --train-global-batch-size 1024 will accumulate gradients over 8 iterations before doing the weight update with accumulated gradients.

6. Start evaluation.

The training process automatically runs evaluation and outputs the BLEU score after each training epoch. Additionally, after the training is done, you can manually run inference on the test dataset with the checkpoint saved during the training.

To launch mixed precision inference on the newstest2014.en test set, run:

python3 translate.py \
  --input data/wmt16_de_en/newstest2014.en \
  --reference data/wmt16_de_en/newstest2014.de \
  --output /tmp/output \
  --model results/gnmt/model_best.pth

The script will load the checkpoint specified by the --model option, then it will launch inference on the file specified by the --input option, and compute BLEU score against the reference translation specified by the --reference option. Outputs will be stored to the location specified by the --output option.

Additionally, one can pass the input text directly from the command-line:

python3 translate.py \
  --input-text "The quick brown fox jumps over the lazy dog" \
  --model results/gnmt/model_best.pth

Translated output will be printed to the console:

(...)
0: Translated output:
Der schnelle braune Fuchs springt über den faulen Hund

By default, the translate.py script will launch mixed precision inference with Tensor Cores. You can change this behavior by setting the --math fp32 flag for the translate.py inference script.

Advanced

The following sections provide greater details of the dataset, running training and inference, and the training results.

Scripts and sample code

In the root directory, the most important files are:

• train.py: serves as the entry point to launch the training
• translate.py: serves as the entry point to launch inference
• Dockerfile: container with the basic set of dependencies to run GNMT v2
• requirements.txt: set of extra requirements for running GNMT v2

The seq2seq/model directory contains the implementation of GNMT v2 building blocks:

• attention.py: implementation of normalized Bahdanau attention
• encoder.py: implementation of recurrent encoder
• decoder.py: implementation of recurrent decoder with attention
• seq2seq_base.py: base class for seq2seq models
• gnmt.py: implementation of GNMT v2 model

The seq2seq/train directory encapsulates the necessary tools to execute training:

• trainer.py: implementation of training loop
• smoothing.py: implementation of cross-entropy with label smoothing
• lr_scheduler.py: implementation of exponential learning rate warmup and step decay
• fp_optimizers.py: implementation of optimizers for various floating point precisions

The seq2seq/inference directory contains scripts required to run inference:

• beam_search.py: implementation of beam search with length normalization and length penalty
• translator.py: implementation of auto-regressive inference

The seq2seq/data directory contains implementation of components needed for data loading:

• dataset.py: implementation of text datasets
• sampler.py: implementation of batch samplers with bucketing by sequence length
• tokenizer.py: implementation of tokenizer (maps integer vocabulary indices to text)

Parameters

Training

The complete list of available parameters for the train.py training script contains:

dataset setup:
  --dataset-dir DATASET_DIR
                        path to the directory with training/test data
                        (default: data/wmt16_de_en)
  --src-lang SRC_LANG   source language (default: en)
  --tgt-lang TGT_LANG   target language (default: de)
  --vocab VOCAB         path to the vocabulary file (relative to DATASET_DIR
                        directory) (default: vocab.bpe.32000)
  -bpe BPE_CODES, --bpe-codes BPE_CODES
                        path to the file with bpe codes (relative to
                        DATASET_DIR directory) (default: bpe.32000)
  --train-src TRAIN_SRC
                        path to the training source data file (relative to
                        DATASET_DIR directory) (default:
                        train.tok.clean.bpe.32000.en)
  --train-tgt TRAIN_TGT
                        path to the training target data file (relative to
                        DATASET_DIR directory) (default:
                        train.tok.clean.bpe.32000.de)
  --val-src VAL_SRC     path to the validation source data file (relative to
                        DATASET_DIR directory) (default:
                        newstest_dev.tok.clean.bpe.32000.en)
  --val-tgt VAL_TGT     path to the validation target data file (relative to
                        DATASET_DIR directory) (default:
                        newstest_dev.tok.clean.bpe.32000.de)
  --test-src TEST_SRC   path to the test source data file (relative to
                        DATASET_DIR directory) (default:
                        newstest2014.tok.bpe.32000.en)
  --test-tgt TEST_TGT   path to the test target data file (relative to
                        DATASET_DIR directory) (default: newstest2014.de)
  --train-max-size TRAIN_MAX_SIZE
                        use at most TRAIN_MAX_SIZE elements from training
                        dataset (useful for benchmarking), by default uses
                        entire dataset (default: None)

results setup:
  --results-dir RESULTS_DIR
                        path to directory with results, it will be
                        automatically created if it does not exist (default:
                        results)
  --save-dir SAVE_DIR   defines subdirectory within RESULTS_DIR for results
                        from this training run (default: gnmt)
  --print-freq PRINT_FREQ
                        print log every PRINT_FREQ batches (default: 10)

model setup:
  --hidden-size HIDDEN_SIZE
                        hidden size of the model (default: 1024)
  --num-layers NUM_LAYERS
                        number of RNN layers in encoder and in decoder
                        (default: 4)
  --dropout DROPOUT     dropout applied to input of RNN cells (default: 0.2)
  --share-embedding     use shared embeddings for encoder and decoder (use '--
                        no-share-embedding' to disable) (default: True)
  --smoothing SMOOTHING
                        label smoothing, if equal to zero model will use
                        CrossEntropyLoss, if not zero model will be trained
                        with label smoothing loss (default: 0.1)

general setup:
  --math {fp16,fp32,manual_fp16}
                        precision (default: fp16)
  --seed SEED           master seed for random number generators, if "seed" is
                        undefined then the master seed will be sampled from
                        random.SystemRandom() (default: None)
  --prealloc-mode {off,once,always}
                        controls preallocation (default: always)
  --eval                run validation and test after every epoch (use '--no-
                        eval' to disable) (default: True)
  --env                 print info about execution env (use '--no-env' to
                        disable) (default: True)
  --cuda                enables cuda (use '--no-cuda' to disable) (default:
                        True)
  --cudnn               enables cudnn (use '--no-cudnn' to disable) (default:
                        True)
  --log-all-ranks       enables logging from all distributed ranks, if
                        disabled then only logs from rank 0 are reported (use
                        '--no-log-all-ranks' to disable) (default: True)

training setup:
  --train-batch-size TRAIN_BATCH_SIZE
                        training batch size per worker (default: 128)
  --train-global-batch-size TRAIN_GLOBAL_BATCH_SIZE
                        global training batch size, this argument does not
                        have to be defined, if it is defined it will be used
                        to automatically compute train_iter_size using the
                        equation: train_iter_size = train_global_batch_size //
                        (train_batch_size * world_size) (default: None)
  --train-iter-size N   training iter size, training loop will accumulate
                        gradients over N iterations and execute optimizer
                        every N steps (default: 1)
  --epochs EPOCHS       max number of training epochs (default: 6)
  --grad-clip GRAD_CLIP
                        enables gradient clipping and sets maximum norm of
                        gradients (default: 5.0)
  --train-max-length TRAIN_MAX_LENGTH
                        maximum sequence length for training (including
                        special BOS and EOS tokens) (default: 50)
  --train-min-length TRAIN_MIN_LENGTH
                        minimum sequence length for training (including
                        special BOS and EOS tokens) (default: 0)
  --train-loader-workers TRAIN_LOADER_WORKERS
                        number of workers for training data loading (default:
                        2)
  --batching {random,sharding,bucketing}
                        select batching algorithm (default: bucketing)
  --shard-size SHARD_SIZE
                        shard size for "sharding" batching algorithm, in
                        multiples of global batch size (default: 80)
  --num-buckets NUM_BUCKETS
                        number of buckets for "bucketing" batching algorithm
                        (default: 5)

optimizer setup:
  --optimizer OPTIMIZER
                        training optimizer (default: SparseAdam)
  --lr LR               learning rate (default: 0.002)
  --optimizer-extra OPTIMIZER_EXTRA
                        extra options for the optimizer (default: {})

mixed precision loss scaling setup:
  --init-scale INIT_SCALE
                        initial loss scale (default: 8192)
  --upscale-interval UPSCALE_INTERVAL
                        loss upscaling interval (default: 128)

learning rate scheduler setup:
  --warmup-steps WARMUP_STEPS
                        number of learning rate warmup iterations (default:
                        200)
  --remain-steps REMAIN_STEPS
                        starting iteration for learning rate decay (default:
                        0.666)
  --decay-interval DECAY_INTERVAL
                        interval between learning rate decay steps (default:
                        None)
  --decay-steps DECAY_STEPS
                        max number of learning rate decay steps (default: 4)
  --decay-factor DECAY_FACTOR
                        learning rate decay factor (default: 0.5)

validation setup:
  --val-batch-size VAL_BATCH_SIZE
                        batch size for validation (default: 64)
  --val-max-length VAL_MAX_LENGTH
                        maximum sequence length for validation (including
                        special BOS and EOS tokens) (default: 125)
  --val-min-length VAL_MIN_LENGTH
                        minimum sequence length for validation (including
                        special BOS and EOS tokens) (default: 0)
  --val-loader-workers VAL_LOADER_WORKERS
                        number of workers for validation data loading
                        (default: 0)

test setup:
  --test-batch-size TEST_BATCH_SIZE
                        batch size for test (default: 128)
  --test-max-length TEST_MAX_LENGTH
                        maximum sequence length for test (including special
                        BOS and EOS tokens) (default: 150)
  --test-min-length TEST_MIN_LENGTH
                        minimum sequence length for test (including special
                        BOS and EOS tokens) (default: 0)
  --beam-size BEAM_SIZE
                        beam size (default: 5)
  --len-norm-factor LEN_NORM_FACTOR
                        length normalization factor (default: 0.6)
  --cov-penalty-factor COV_PENALTY_FACTOR
                        coverage penalty factor (default: 0.1)
  --len-norm-const LEN_NORM_CONST
                        length normalization constant (default: 5.0)
  --intra-epoch-eval N  evaluate within training epoch, this option will
                        enable extra N equally spaced evaluations executed
                        during each training epoch (default: 0)
  --test-loader-workers TEST_LOADER_WORKERS
                        number of workers for test data loading (default: 0)

checkpointing setup:
  --start-epoch START_EPOCH
                        manually set initial epoch counter (default: 0)
  --resume PATH         resumes training from checkpoint from PATH (default:
                        None)
  --save-all            saves checkpoint after every epoch (default: False)
  --save-freq SAVE_FREQ
                        save checkpoint every SAVE_FREQ batches (default:
                        5000)
  --keep-checkpoints KEEP_CHECKPOINTS
                        keep only last KEEP_CHECKPOINTS checkpoints, affects
                        only checkpoints controlled by --save-freq option
                        (default: 0)

benchmark setup:
  --target-perf TARGET_PERF
                        target training performance (in tokens per second)
                        (default: None)
  --target-bleu TARGET_BLEU
                        target accuracy (default: None)

Inference

The complete list of available parameters for the translate.py inference script contains:

data setup:
  -o OUTPUT, --output OUTPUT
                        full path to the output file if not specified, then
                        the output will be printed (default: None)
  -r REFERENCE, --reference REFERENCE
                        full path to the file with reference translations (for
                        sacrebleu, raw text) (default: None)
  -m MODEL, --model MODEL
                        full path to the model checkpoint file (default: None)
  -i INPUT, --input INPUT
                        full path to the input file (raw text) (default: None)
  -t INPUT_TEXT [INPUT_TEXT ...], --input-text INPUT_TEXT [INPUT_TEXT ...]
                        raw input text (default: None)
  --sort                sorts dataset by sequence length (use '--no-sort' to
                        disable) (default: False)

inference setup:
  --batch-size BATCH_SIZE [BATCH_SIZE ...]
                        batch size per GPU (default: [128])
  --beam-size BEAM_SIZE [BEAM_SIZE ...]
                        beam size (default: [5])
  --max-seq-len MAX_SEQ_LEN
                        maximum generated sequence length (default: 80)
  --len-norm-factor LEN_NORM_FACTOR
                        length normalization factor (default: 0.6)
  --cov-penalty-factor COV_PENALTY_FACTOR
                        coverage penalty factor (default: 0.1)
  --len-norm-const LEN_NORM_CONST
                        length normalization constant (default: 5.0)

general setup:
  --math {fp16,fp32} [{fp16,fp32} ...]
                        arithmetic type (default: ['fp16'])
  --env                 print info about execution env (use '--no-env' to
                        disable) (default: False)
  --bleu                compares with reference translation and computes BLEU
                        (use '--no-bleu' to disable) (default: True)
  --cuda                enables cuda (use '--no-cuda' to disable) (default:
                        True)
  --cudnn               enables cudnn (use '--no-cudnn' to disable) (default:
                        True)
  --batch-first         uses (batch, seq, feature) data format for RNNs
                        (default: True)
  --seq-first           uses (seq, batch, feature) data format for RNNs
                        (default: True)
  --print-freq PRINT_FREQ, -p PRINT_FREQ
                        print log every PRINT_FREQ batches (default: 1)

benchmark setup:
  --target-perf TARGET_PERF
                        target inference performance (in tokens per second)
                        (default: None)
  --target-bleu TARGET_BLEU
                        target accuracy (default: None)
  --repeat REPEAT [REPEAT ...]
                        loops over the dataset REPEAT times, flag accepts
                        multiple arguments, one for each specified batch size
                        (default: [1])
  --warmup WARMUP       warmup iterations for performance counters (default:
                        0)
  --percentiles PERCENTILES [PERCENTILES ...]
                        Percentiles for confidence intervals for
                        throughput/latency benchmarks (default: (50, 90, 95,
                        99, 100))
  --tables              print accuracy, throughput and latency results in
                        tables (use '--no-tables' to disable) (default: False)

Command-line options

To see the full list of available options and their descriptions, use the -h or --help command line option. For example, for training:

python3 train.py --help

usage: train.py [-h] [--dataset-dir DATASET_DIR] [--src-lang SRC_LANG]
                [--tgt-lang TGT_LANG] [--vocab VOCAB] [-bpe BPE_CODES]
                [--train-src TRAIN_SRC] [--train-tgt TRAIN_TGT]
                [--val-src VAL_SRC] [--val-tgt VAL_TGT] [--test-src TEST_SRC]
                [--test-tgt TEST_TGT] [--results-dir RESULTS_DIR]
                [--save-dir SAVE_DIR] [--print-freq PRINT_FREQ]
                [--hidden-size HIDDEN_SIZE] [--num-layers NUM_LAYERS]
                [--dropout DROPOUT] [--share-embedding]
                [--smoothing SMOOTHING] [--math {fp16,fp32,manual_fp16}]
                [--seed SEED] [--prealloc-mode {off,once,always}] [--eval]
                [--env] [--cuda] [--cudnn] [--log-all-ranks]
                [--train-max-size TRAIN_MAX_SIZE]
                [--train-batch-size TRAIN_BATCH_SIZE]
                [--train-global-batch-size TRAIN_GLOBAL_BATCH_SIZE]
                [--train-iter-size N] [--epochs EPOCHS]
                [--grad-clip GRAD_CLIP] [--train-max-length TRAIN_MAX_LENGTH]
                [--train-min-length TRAIN_MIN_LENGTH]
                [--train-loader-workers TRAIN_LOADER_WORKERS]
                [--batching {random,sharding,bucketing}]
                [--shard-size SHARD_SIZE] [--num-buckets NUM_BUCKETS]
                [--optimizer OPTIMIZER] [--lr LR]
                [--optimizer-extra OPTIMIZER_EXTRA] [--init-scale INIT_SCALE]
                [--upscale-interval UPSCALE_INTERVAL]
                [--warmup-steps WARMUP_STEPS] [--remain-steps REMAIN_STEPS]
                [--decay-interval DECAY_INTERVAL] [--decay-steps DECAY_STEPS]
                [--decay-factor DECAY_FACTOR]
                [--val-batch-size VAL_BATCH_SIZE]
                [--val-max-length VAL_MAX_LENGTH]
                [--val-min-length VAL_MIN_LENGTH]
                [--val-loader-workers VAL_LOADER_WORKERS]
                [--test-batch-size TEST_BATCH_SIZE]
                [--test-max-length TEST_MAX_LENGTH]
                [--test-min-length TEST_MIN_LENGTH] [--beam-size BEAM_SIZE]
                [--len-norm-factor LEN_NORM_FACTOR]
                [--cov-penalty-factor COV_PENALTY_FACTOR]
                [--len-norm-const LEN_NORM_CONST] [--intra-epoch-eval N]
                [--test-loader-workers TEST_LOADER_WORKERS]
                [--start-epoch START_EPOCH] [--resume PATH] [--save-all]
                [--save-freq SAVE_FREQ] [--keep-checkpoints KEEP_CHECKPOINTS]
                [--target-perf TARGET_PERF] [--target-bleu TARGET_BLEU]
                [--rank RANK] [--local_rank LOCAL_RANK]

For example, for inference:

python3 translate.py --help

usage: translate.py [-h] [-o OUTPUT] [-r REFERENCE] -m MODEL
                    (-i INPUT | -t INPUT_TEXT [INPUT_TEXT ...]) [--sort]
                    [--batch-size BATCH_SIZE [BATCH_SIZE ...]]
                    [--beam-size BEAM_SIZE [BEAM_SIZE ...]]
                    [--max-seq-len MAX_SEQ_LEN]
                    [--len-norm-factor LEN_NORM_FACTOR]
                    [--cov-penalty-factor COV_PENALTY_FACTOR]
                    [--len-norm-const LEN_NORM_CONST]
                    [--math {fp16,fp32} [{fp16,fp32} ...]] [--env] [--bleu]
                    [--cuda] [--cudnn] [--batch-first | --seq-first]
                    [--print-freq PRINT_FREQ] [--target-perf TARGET_PERF]
                    [--target-bleu TARGET_BLEU] [--repeat REPEAT [REPEAT ...]]
                    [--warmup WARMUP]
                    [--percentiles PERCENTILES [PERCENTILES ...]] [--tables]
                    [--rank RANK] [--local_rank LOCAL_RANK]

Getting the data

The GNMT v2 model was trained on the WMT16 English-German dataset. Concatenation of the newstest2015 and newstest2016 test sets are used as a validation dataset and the newstest2014 is used as a testing dataset.

This repository contains the scripts/wmt16_en_de.sh download script which automatically downloads and preprocesses the training, validation and test datasets. By default, data is downloaded to the data directory.

Our download script is very similar to the wmt16_en_de.sh script from the tensorflow/nmt repository. Our download script contains an extra preprocessing step, which discards all pairs of sentences which can't be decoded by latin-1 encoder. The scripts/wmt16_en_de.sh script uses the subword-nmt package to segment text into subword units (Byte Pair Encodings - BPE). By default, the script builds the shared vocabulary of 32,000 tokens.

In order to test with other datasets, the script needs to be customized accordingly.

Dataset guidelines

The process of downloading and preprocessing the data can be found in the scripts/wmt16_en_de.sh script.

Initially, data is downloaded from www.statmt.org. Then europarl-v7, commoncrawl and news-commentary corpora are concatenated to form the training dataset, similarly newstest2015 and newstest2016 are concatenated to form the validation dataset. Raw data is preprocessed with Moses, first by launching Moses tokenizer (tokenizer breaks up text into individual words), then by launching clean-corpus-n.perl which removes invalid sentences and does initial filtering by sequence length.

Second stage of preprocessing is done by launching the scripts/filter_dataset.py script, which discards all pairs of sentences that can't be decoded by latin-1 encoder.

Third state of preprocessing uses the subword-nmt package. First it builds shared byte pair encoding vocabulary with 32,000 merge operations (command subword-nmt learn-bpe), then it applies generated vocabulary to training, validation and test corpora (command subword-nmt apply-bpe).

Training process

The default training configuration can be launched by running the train.py training script. By default, the training script saves only one checkpoint with the lowest value of the loss function on the validation dataset. An evaluation is then performed after each training epoch. Results are stored in the results/gnmt directory.

The training script launches data-parallel training with batch size 128 per GPU on all available GPUs. We have tested reliance on up to 16 GPUs on a single node. After each training epoch, the script runs an evaluation on the validation dataset and outputs a BLEU score on the test dataset (newstest2014). BLEU is computed by the SacreBLEU package. Logs from the training and evaluation are saved to the results directory.

The summary after each training epoch is printed in the following format:

0: Summary: Epoch: 3	Training Loss: 3.1336	Validation Loss: 2.9587	Test BLEU: 23.18
0: Performance: Epoch: 3	Training: 418772 Tok/s	Validation: 1445331 Tok/s

The training loss is averaged over an entire training epoch, the validation loss is averaged over the validation dataset and the BLEU score is computed on the test dataset. Performance is reported in total tokens per second. The result is averaged over an entire training epoch and summed over all GPUs participating in the training.

Even though the training script uses all available GPUs, you can change this behavior by setting the CUDA_VISIBLE_DEVICES variable in your environment or by setting the NV_GPU variable at the Docker container launch (see section "GPU isolation").

By default, the train.py script will launch mixed precision training with Tensor Cores. You can change this behavior by setting the --math fp32 flag for the train.py script.

To view all available options for training, run python3 train.py --help.

Inference process

Inference can be run by launching the translate.py inference script, although, it requires a pre-trained model checkpoint and tokenized input.

The inference script, translate.py, supports batched inference. By default, it launches beam search with beam size of 5, coverage penalty term and length normalization term. Greedy decoding can be enabled by setting the beam size to 1.

To view all available options for inference, run python3 translate.py --help.

Performance

Benchmarking

The following section shows how to run benchmarks measuring the model performance in training and inference modes.

Training performance benchmark

Training is launched on batches of text data, different batches have different sequence lengths (number of tokens in the longest sequence). Sequence length and batch efficiency (ratio of non-pad tokens to total number of tokens) affect performance of the training, therefore it's recommended to run the training on a large chunk of training dataset to get a stable and reliable average training performance. Ideally at least one full epoch of training should be launched to get a good estimate of training performance.

The following commands will launch one epoch of training:

To launch mixed precision training on 1, 4 or 8 GPUs, run:

python3 -m launch train.py --seed 2 --train-global-batch-size 1024 --epochs 1 --math fp16

To launch mixed precision training on 16 GPUs, run:

python3 -m launch train.py --seed 2 --train-global-batch-size 2048 --epochs 1 --math fp16

Change --math fp16 to --math fp32 to launch a single precision training.

After the training is completed, the train.py script prints a summary to standard output. Performance results are printed in the following format:

(...)
0: Performance: Epoch: 0	Training: 418926 Tok/s	Validation: 1430828 Tok/s
(...)

Training: 418926 Tok/s represents training throughput averaged over an entire training epoch and summed over all GPUs participating in the training.

Inference performance benchmark

The inference performance and accuracy benchmarks require a checkpoint from a fully trained model.

Command to launch the inference accuracy benchmark:

python3 translate.py \
  --model results/gnmt/model_best.pth \
  --input data/wmt16_de_en/newstest2014.en \
  --reference data/wmt16_de_en/newstest2014.de \
  --output /tmp/output \
  --math fp16 fp32 \
  --batch-size 128 \
  --beam-size 1 2 5 \
  --tables

Command to launch the inference throughput and latency benchmarks:

python3 translate.py \
  --model results/gnmt/model_best.pth \
  --input data/wmt16_de_en/newstest2014.en \
  --reference data/wmt16_de_en/newstest2014.de \
  --output /tmp/output \
  --math fp16 fp32 \
  --batch-size 1 2 4 8 32 128 512 \
  --repeat 1 1 1 1 2 8 16 \
  --beam-size 1 2 5 \
  --warmup 5 \
  --tables

Results

The following sections provide details on how we achieved our performance and accuracy in training and inference.

Training accuracy results

Training accuracy: NVIDIA DGX-1 (8x V100 16G)

Our results were obtained by running the train.py script with the default batch size = 128 per GPU in the pytorch-19.05-py3 NGC container on NVIDIA DGX-1 with (8x V100 16G) GPUs.

Command to launch the training:

python3 -m launch train.py --seed 2 --train-global-batch-size 1024 --math fp16

Change --math fp16 to --math fp32 to launch a single precision training.

GPUs	Batch Size / GPU	Accuracy - FP32 (BLEU)	Accuracy - Mixed precision (BLEU)	Time to Train - FP32 (minutes)	Time to Train - Mixed precision (minutes)	Time to Train Speedup (FP32 to Mixed precision)
1	128	24.41	24.41	821.2	256.0	3.21
4	128	24.43	24.51	232.3	79.0	2.94
8	128	24.45	24.48	118.1	42.5	2.78

To achieve these same results, follow the Quick Start Guide outlined above.

Training accuracy: NVIDIA DGX-2 (16x V100 32G)

Our results were obtained by running the train.py script with the default batch size = 128 per GPU in the pytorch-19.05-py3 NGC container on NVIDIA DGX-2 with (16x V100 32G) GPUs.

Commands to launch the training:

To launch mixed precision training on 1, 4 or 8 GPUs, run:

python3 -m launch train.py --seed 2 --train-global-batch-size 1024 --math fp16

To launch mixed precision training on 16 GPUs, run:

python3 -m launch train.py --seed 2 --train-global-batch-size 2048 --math fp16

Change --math fp16 to --math fp32 to launch a single precision training.

GPUs	Batch Size / GPU	Accuracy - FP32 (BLEU)	Accuracy - Mixed precision (BLEU)	Time to Train - FP32 (minutes)	Time to Train - Mixed precision (minutes)	Time to Train Speedup (FP32 to Mixed precision)
1	128	24.41	24.41	831.4	240.8	3.45
4	128	24.43	24.45	219.2	74.5	2.94
8	128	24.72	24.56	114.6	42.4	2.70
16	128	23.98	24.08	59.1	23.3	2.54

To achieve these same results, follow the Quick Start Guide outlined above.

Training stability test

The GNMT v2 model was trained for 6 epochs, starting from 50 different initial random seeds. After each training epoch, the model was evaluated on the test dataset and the BLEU score was recorded. The training was performed in the pytorch-19.05-py3 Docker container on NVIDIA DGX-1 with 8 Tesla V100 16G GPUs. The following table summarizes the results of the stability test.

In the following table, the BLEU scores after each training epoch for different initial random seeds are displayed.

Epoch	Average	Standard deviation	Minimum	Maximum	Median
1	19.960	0.347	18.460	20.510	19.975
2	21.778	0.248	21.190	22.170	21.790
3	22.501	0.210	21.890	22.870	22.475
4	23.148	0.169	22.660	23.480	23.165
5	24.158	0.140	23.910	24.460	24.155
6	24.378	0.165	24.010	24.690	24.395

Training throughput results

Training throughput: NVIDIA DGX-1 (8x V100 16G)

Our results were obtained by running the train.py training script in the pytorch-19.05-py3 NGC container on NVIDIA DGX-1 with (8x V100 16G) GPUs. Throughput performance numbers (in tokens per second) were averaged over an entire training epoch.

GPUs	Batch size / GPU	Throughput - FP32 (tok/s)	Throughput - Mixed precision (tok/s)	Throughput speedup (FP32 to Mixed precision)	Strong Scaling - FP32	Strong Scaling - Mixed precision
1	128	21424	68312	3.189	1.000	1.000
4	128	75658	221308	2.925	3.531	3.240
8	128	149552	419075	2.802	6.981	6.135

To achieve these same results, follow the Quick Start Guide outlined above.

Training throughput: NVIDIA DGX-2 (16x V100 32G)

Our results were obtained by running the train.py training script in the pytorch-19.05-py3 NGC container on NVIDIA DGX-2 with (16x V100 32G) GPUs. Throughput performance numbers (in tokens per second) were averaged over an entire training epoch.

GPUs	Batch size / GPU	Throughput - FP32 (tok/s)	Throughput - Mixed precision (tok/s)	Throughput speedup (FP32 to Mixed precision)	Strong Scaling - FP32	Strong Scaling - Mixed precision
1	128	22742	72684	3.196	1.000	1.000
4	128	80395	237616	2.956	3.535	3.269
8	128	155297	430377	2.771	6.829	5.921
16	128	312426	852550	2.729	13.738	11.730

To achieve these same results, follow the Quick Start Guide outlined above.

Inference accuracy results

Inference accuracy: NVIDIA Tesla V100 16G

Our results were obtained by running the translate.py script in the pytorch-19.05-py3 NGC Docker container with NVIDIA Tesla V100 16G GPUs. Full command to launch the inference accuracy benchmark was provided in the Inference performance benchmark section.

Batch Size	Beam Size	Accuracy - FP32 (BLEU)	Accuracy - FP16 (BLEU)
128	1	23.07	23.07
128	2	23.81	23.79
128	5	24.40	24.43

Inference throughput results

Tables presented in this section show the average inference throughput (columns Avg (tok/s)) and inference throughput for various confidence intervals (columns N% (ms), where N denotes the confidence interval). Inference throughput is measured in tokens per second. Speedups reported in FP16 subsections are relative to FP32 numbers for corresponding configuration.

Inference throughput: NVIDIA Tesla V100 16G

Our results were obtained by running the translate.py script in the pytorch-19.05-py3 NGC Docker container with NVIDIA Tesla V100 16G GPUs. Full command to launch the inference throughput benchmark was provided in the Inference performance benchmark section.

FP32

Batch Size	Beam Size	Avg (tok/s)	50% (tok/s)	90% (tok/s)	95% (tok/s)	99% (tok/s)	100% (tok/s)
1	1	936.7	930.9	869.9	851.5	811.5	695.7
1	2	640.5	642.7	584.6	561.1	512.1	393.6
1	5	595.0	600.9	531.3	503.2	456.7	268.5
2	1	1452.3	1455.9	1204.7	1145.1	1060.7	969.5
2	2	1032.5	1038.5	851.2	808.5	746.5	642.4
2	5	979.5	981.9	806.5	773.1	701.5	601.4
4	1	2425.7	2414.3	1990.4	1876.3	1706.3	1505.0
4	2	1697.6	1690.2	1401.5	1325.1	1180.5	989.9
4	5	1634.4	1631.5	1344.3	1274.0	1138.3	993.9
8	1	4162.5	4176.7	3462.9	3254.8	2885.4	2575.2
8	2	2953.0	2961.9	2438.4	2288.9	2073.9	1817.2
8	5	2723.0	2716.6	2299.6	2138.5	1888.9	1855.2
32	1	12178.6	12181.0	10445.3	9805.9	9253.3	8584.7
32	2	7990.9	8067.5	6785.3	6354.9	5912.3	5900.5
32	5	5640.1	5650.5	4951.4	4771.7	4486.0	4446.6
128	1	25192.7	25083.5	22951.1	20610.8	20068.7	19750.6
128	2	14639.3	14662.0	13285.7	12482.4	12100.9	12031.9
128	5	8610.5	8607.0	7880.1	7589.0	7129.6	7069.2
512	1	33281.8	33442.0	29899.4	29721.3	29644.1	29568.4
512	2	19803.8	19867.4	17872.7	17757.6	17693.8	17669.8
512	5	9236.4	9282.9	8309.0	8272.7	8241.2	8228.6

FP16

Batch Size	Beam Size	Avg (tok/s)	Speedup	50% (tok/s)	Speedup	90% (tok/s)	Speedup	95% (tok/s)	Speedup	99% (tok/s)	Speedup	100% (tok/s)	Speedup
1	1	920.8	0.9831	917.6	0.9857	844.5	0.9708	820.0	0.9630	751.8	0.9264	503.5	0.7238
1	2	623.7	0.9738	625.7	0.9736	563.5	0.9639	539.1	0.9609	485.4	0.9479	389.4	0.9894
1	5	597.3	1.0038	603.9	1.0049	531.1	0.9996	501.2	0.9960	450.8	0.9869	260.0	0.9684
2	1	1496.5	1.0305	1499.2	1.0298	1228.8	1.0201	1165.6	1.0179	1074.5	1.0130	936.7	0.9661
2	2	1027.5	0.9951	1033.2	0.9949	848.6	0.9970	801.2	0.9910	737.5	0.9879	644.7	1.0036
2	5	994.5	1.0153	994.1	1.0124	815.1	1.0106	777.2	1.0052	710.4	1.0126	605.8	1.0073
4	1	2500.6	1.0309	2498.7	1.0349	2038.1	1.0240	1913.4	1.0197	1722.5	1.0095	1534.9	1.0199
4	2	1707.7	1.0060	1697.5	1.0043	1402.2	1.0005	1330.9	1.0043	1198.9	1.0157	986.5	0.9965
4	5	1668.5	1.0209	1664.2	1.0201	1367.9	1.0176	1287.7	1.0107	1165.0	1.0235	969.7	0.9757
8	1	4489.7	1.0786	4498.1	1.0769	3693.0	1.0664	3451.2	1.0603	3048.2	1.0564	2667.5	1.0358
8	2	3025.4	1.0245	3026.8	1.0219	2462.6	1.0099	2295.7	1.0029	2108.3	1.0166	1877.5	1.0332
8	5	2925.9	1.0745	2920.0	1.0749	2433.6	1.0583	2302.0	1.0765	1996.1	1.0568	1890.2	1.0189
32	1	13953.9	1.1458	13976.1	1.1474	11972.1	1.1462	11480.7	1.1708	10203.9	1.1027	9704.5	1.1304
32	2	9095.4	1.1382	9206.5	1.1412	7718.9	1.1376	7344.8	1.1558	6721.9	1.1369	6620.4	1.1220
32	5	8556.1	1.5170	8610.8	1.5239	7098.3	1.4336	6860.5	1.4377	6511.3	1.4515	6371.9	1.4330
128	1	39354.1	1.5621	39356.0	1.5690	35942.3	1.5660	34173.7	1.6580	29650.9	1.4775	29355.0	1.4863
128	2	24597.0	1.6802	24458.4	1.6682	22311.8	1.6794	21035.5	1.6852	19444.6	1.6069	19336.3	1.6071
128	5	18740.1	2.1764	18698.6	2.1725	17010.1	2.1586	15965.2	2.1037	15454.6	2.1677	15219.2	2.1529
512	1	78936.4	2.3718	79001.1	2.3623	73298.5	2.4515	72552.1	2.4411	71825.3	2.4229	71451.9	2.4165
512	2	47856.9	2.4166	47924.3	2.4122	44186.7	2.4723	44055.8	2.4810	43574.9	2.4627	43425.8	2.4576
512	5	28243.8	3.0579	28268.4	3.0452	25709.9	3.0942	25604.0	3.0950	25373.4	3.0789	25196.4	3.0621

To achieve these same results, follow the Quick Start Guide outlined above.

Inference latency results

Tables presented in this section show the average inference latency (columns Avg (ms)) and inference latency for various confidence intervals (columns N% (ms), where N denotes the confidence interval). Inference latency is measured in milliseconds. Speedups reported in FP16 subsections are relative to FP32 numbers for corresponding configuration.

Inference latency: NVIDIA Tesla V100 16G

Our results were obtained by running the translate.py script in the pytorch-19.05-py3 NGC Docker container with NVIDIA Tesla V100 16G GPUs. Full command to launch the inference latency benchmark was provided in the Inference performance benchmark section.

FP32

Batch Size	Beam Size	Avg (ms)	50% (ms)	90% (ms)	95% (ms)	99% (ms)	100% (ms)
1	1	61.90	56.96	103.4	117.9	139.2	171.2
1	2	89.38	82.67	145.9	165.1	194.8	239.3
1	5	95.82	88.84	154.9	175.4	206.7	253.3
2	1	80.29	76.37	121.7	132.0	152.8	179.0
2	2	112.48	106.88	167.3	181.1	211.5	251.3
2	5	118.31	112.14	174.8	191.2	225.5	253.8
4	1	96.19	94.35	132.3	142.6	164.4	183.0
4	2	137.06	134.90	186.2	201.1	234.8	255.8
4	5	142.28	140.10	195.3	207.0	243.9	255.8
8	1	111.78	110.94	144.9	153.6	173.8	191.2
8	2	157.16	157.23	200.9	213.3	243.0	254.3
8	5	170.33	169.87	215.2	235.9	265.0	273.5
32	1	152.15	152.61	183.2	193.3	201.2	210.9
32	2	232.17	230.32	283.0	294.7	307.4	327.8
32	5	328.42	328.79	397.2	417.2	431.4	437.4
128	1	292.81	294.50	322.3	328.2	360.0	361.0
128	2	504.35	504.82	556.9	574.8	610.8	619.2
128	5	857.25	863.34	956.6	980.6	1045.4	1062.1
512	1	885.94	868.39	1002.8	1008.1	1016.9	1017.4
512	2	1490.94	1476.64	1683.1	1697.5	1705.0	1709.8
512	5	3195.79	3168.53	3596.3	3641.5	3656.5	3671.3

FP16

Batch Size	Beam Size	Avg (ms)	Speedup	50% (ms)	Speedup	90% (ms)	Speedup	95% (ms)	Speedup	99% (ms)	Speedup	100% (ms)	Speedup
1	1	62.91	0.9840	58.30	0.9770	104.4	0.9898	119.6	0.9858	143.5	0.9701	176.9	0.9677
1	2	91.97	0.9718	84.61	0.9771	150.1	0.9717	169.3	0.9756	199.9	0.9747	260.2	0.9198
1	5	95.55	1.0029	88.14	1.0079	155.1	0.9989	175.5	0.9991	209.7	0.9860	252.9	1.0015
2	1	78.18	1.0270	74.05	1.0313	118.7	1.0252	129.2	1.0219	151.6	1.0079	181.9	0.9843
2	2	113.11	0.9944	107.32	0.9959	168.6	0.9926	181.9	0.9953	211.6	0.9995	249.8	1.0062
2	5	116.67	1.0141	110.88	1.0114	174.1	1.0040	188.3	1.0156	221.5	1.0181	246.6	1.0290
4	1	93.53	1.0284	90.98	1.0370	129.0	1.0254	140.8	1.0121	159.7	1.0289	175.4	1.0433
4	2	136.42	1.0047	134.75	1.0012	184.6	1.0087	200.0	1.0052	235.1	0.9989	247.9	1.0319
4	5	139.49	1.0200	136.88	1.0236	191.1	1.0219	202.3	1.0230	240.2	1.0153	257.1	0.9948
8	1	103.75	1.0774	103.01	1.0769	135.6	1.0682	144.4	1.0635	162.1	1.0724	168.6	1.1338
8	2	153.57	1.0234	152.57	1.0306	197.3	1.0184	209.8	1.0167	236.7	1.0268	244.4	1.0405
8	5	158.68	1.0734	157.76	1.0768	200.3	1.0745	215.8	1.0934	245.0	1.0815	246.7	1.1087
32	1	132.89	1.1449	133.35	1.1444	160.5	1.1414	167.4	1.1548	173.9	1.1570	176.9	1.1922
32	2	203.80	1.1392	202.76	1.1359	245.3	1.1540	254.7	1.1572	264.9	1.1606	269.8	1.2150
32	5	216.72	1.5154	216.19	1.5208	261.7	1.5182	273.6	1.5247	282.4	1.5276	292.8	1.4940
128	1	187.34	1.5630	188.53	1.5621	204.9	1.5730	208.2	1.5765	213.6	1.6858	219.7	1.6432
128	2	300.33	1.6793	300.59	1.6794	333.2	1.6716	340.8	1.6864	345.2	1.7695	373.7	1.6572
128	5	393.62	2.1778	393.64	2.1932	435.0	2.1993	443.0	2.2136	457.6	2.2847	468.2	2.2686
512	1	373.13	2.3744	367.21	2.3648	407.2	2.4628	410.3	2.4573	414.3	2.4548	414.7	2.4531
512	2	616.52	2.4183	610.26	2.4197	676.7	2.4872	682.2	2.4882	696.0	2.4496	698.5	2.4478
512	5	1044.46	3.0597	1039.28	3.0488	1169.8	3.0743	1172.5	3.1056	1180.9	3.0963	1183.9	3.1011

To achieve these same results, follow the Quick Start Guide outlined above.

Release notes

Changelog

1. Aug 7, 2018

• Initial release

2. Dec 4, 2018

• Added exponential warm-up and step learning rate decay
• Multi-GPU (distributed) inference and validation
• Default container updated to NGC PyTorch 18.11-py3
• General performance improvements

3. Feb 14, 2019

• Different batching algorithm (bucketing with 5 equal-width buckets)
• Additional dropouts before first LSTM layer in encoder and in decoder
• Weight initialization changed to uniform (-0.1,0.1)
• Switched order of dropout and concatenation with attention in decoder
• Default container updated to NGC PyTorch 19.01-py3

4. Jun 25, 2019

• Default container updated to NGC PyTorch 19.05-py3
• Mixed precision training implemented using APEX AMP
• Added inference throughput and latency results on NVIDIA Tesla V100 16G
• Added option to run inference on user-provided raw input text from command line

Known issues

There are no known issues in this release.