Machine Translation Models#

Machine translation is the task of translating text from one language to another. For example, from English to Spanish. Models are based on the Transformer sequence-to-sequence architecture [].

An example script on how to train the model can be found here: NeMo/examples/nlp/machine_translation/enc_dec_nmt.py. The default configuration file for the model can be found at: NeMo/examples/nlp/machine_translation/conf/aayn_base.yaml.

Quick Start Guide#

from nemo.collections.nlp.models import MTEncDecModel

# To get the list of pre-trained models
MTEncDecModel.list_available_models()

# Download and load the a pre-trained to translate from English to Spanish
model = MTEncDecModel.from_pretrained("nmt_en_es_transformer24x6")

# Translate a sentence or list of sentences
translations = model.translate(["Hello!"], source_lang="en", target_lang="es")

Available Models#

Pretrained Models#

Model

Pretrained Checkpoint

New Checkppoints

English -> German

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_de_transformer24x6

German -> English

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_de_en_transformer24x6

English -> Spanish

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_es_transformer24x6

Spanish -> English

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_es_en_transformer24x6

English -> French

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_fr_transformer24x6

French -> English

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_fr_en_transformer24x6

English -> Russian

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_ru_transformer24x6

Russian -> English

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_ru_en_transformer24x6

English -> Chinese

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_zh_transformer24x6

Chinese -> English

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_zh_en_transformer24x6

Old Checkppoints

English -> German

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_de_transformer12x2

German -> English

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_de_en_transformer12x2

English -> Spanish

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_es_transformer12x2

Spanish -> English

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_es_en_transformer12x2

English -> French

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_fr_transformer12x2

French -> English

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_fr_en_transformer12x2

English -> Russian

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_ru_transformer6x6

Russian -> English

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_ru_en_transformer6x6

English -> Chinese

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_zh_transformer6x6

Chinese -> English

https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_zh_en_transformer6x6

Data Format#

Supervised machine translation models require parallel corpora which comprises many examples of sentences in a source language and their corresponding translation in a target language. We use parallel data formatted as separate text files for source and target languages where sentences in corresponding files are aligned like in the table below.

Parallel Coprus#

train.english.txt

train.spanish.txt

Hello .

Hola .

Thank you .

Gracias .

You can now translate from English to Spanish in NeMo .

Ahora puedes traducir del inglés al español en NeMo .

It is common practice to apply data cleaning, normalization, and tokenization to the data prior to training a translation model and NeMo expects already cleaned, normalized, and tokenized data. The only data pre-processing NeMo does is subword tokenization with BPE [].

Data Cleaning, Normalization & Tokenization#

We recommend applying the following steps to clean, normalize, and tokenize your data. All pre-trained models released, apply these data pre-processing steps.

  1. Please take a look at a detailed notebook on best practices to pre-process and clean your datasets - NeMo/tutorials/nlp/Data_Preprocessing_and_Cleaning_for_NMT.ipynb

  2. Language ID filtering - This step filters out examples from your training dataset that aren’t in the correct language. For example, many datasets contain examples where source and target sentences are in the same language. You can use a pre-trained language ID classifier from fastText. Install fastText and then you can then run our script using the lid.176.bin model downloaded from the fastText website.

    python NeMo/scripts/neural_machine_translation/filter_langs_nmt.py \
      --input-src train.en \
      --input-tgt train.es \
      --output-src train_lang_filtered.en \
      --output-tgt train_lang_filtered.es \
      --source-lang en \
      --target-lang es \
      --removed-src train_noise.en \
      --removed-tgt train_noise.es \
      --fasttext-model lid.176.bin
    
  3. Length filtering - We filter out sentences from the data that are below a minimum length (1) or exceed a maximum length (250). We also filter out sentences where the ratio between source and target lengths exceeds 1.3 except for English <-> Chinese models. Moses is a statistical machine translation toolkit that contains many useful pre-processing scripts.

    perl mosesdecoder/scripts/training/clean-corpus-n.perl -ratio 1.3 train en es train.filter 1 250
    
  4. Data cleaning - While language ID filtering can sometimes help with filtering out noisy sentences that contain too many punctuations, it does not help in cases where the translations are potentially incorrect, disfluent, or incomplete. We use bicleaner a tool to identify such sentences. It trains a classifier based on many features included pre-trained language model fluency, word alignment scores from a word-alignment model like Giza++ etc. We use their available pre-trained models wherever possible and train models ourselves using their framework for remaining languages. The following script applies a pre-trained bicleaner model to the data and pick sentences that are clean with probability > 0.5.

    awk '{print "-\t-"}' train.en \
    | paste -d "\t" - train.filter.en train.filter.es \
    | bicleaner-classify - - </path/to/bicleaner.yaml> > train.en-es.bicleaner.score
    
  5. Data deduplication - We use bifixer (which uses xxHash) to hash the source and target sentences based on which we remove duplicate entries from the file. You may want to do something similar to remove training examples that are in the test dataset.

    cat train.en-es.bicleaner.score \
      | parallel -j 25 --pipe -k -l 30000 python bifixer.py --ignore-segmentation -q - - en es \
      > train.en-es.bifixer.score
    
    awk -F awk -F "\t" '!seen[$6]++' train.en-es.bifixer.score > train.en-es.bifixer.dedup.score
    
  6. Filter out data that bifixer assigns probability < 0.5 to.

    awk -F "\t" '{ if ($5>0.5) {print $3}}' train.en-es.bifixer.dedup.score > train.cleaned.en
    awk -F "\t" '{ if ($5>0.5) {print $4}}' train.en-es.bifixer.dedup.score > train.cleaned.es
    
  7. Punctuation Normalization - Punctuation, especially things like quotes can be written in different ways. It’s often useful to normalize the way they appear in text. We use the moses punctuation normalizer on all languages except Chinese.

    perl mosesdecoder/scripts/tokenizer/normalize-punctuation.perl -l es < train.cleaned.es > train.normalized.es
    perl mosesdecoder/scripts/tokenizer/normalize-punctuation.perl -l en < train.cleaned.en > train.normalized.en
    

    For example:

    Before - Aquí se encuentran joyerías como Tiffany`s entre negocios tradicionales suizos como la confitería Sprüngli.
    After  - Aquí se encuentran joyerías como Tiffany's entre negocios tradicionales suizos como la confitería Sprüngli.
    
  8. Tokenization and word segmentation for Chinese - Naturally written text often contains punctuation markers like commas, full-stops and apostrophes that are attached to words. Tokenization by just splitting a string on spaces will result in separate token IDs for very similar items like NeMo and NeMo.. Tokenization splits punctuation from the word to create two separate tokens. In the previous example NeMo. becomes NeMo . which when split by space, results in two tokens and addresses the earlier problem.

    For example:

    Before - Especialmente porque se enfrentará "a Mathieu (Debuchy), Yohan (Cabaye) y Adil (Rami) ", recuerda.
    After  - Especialmente porque se enfrentará " a Mathieu ( Debuchy ) , Yohan ( Cabaye ) y Adil ( Rami ) " , recuerda .
    

    We use the Moses tokenizer for all languages except Chinese.

    perl mosesdecoder/scripts/tokenizer/tokenizer.perl -l es -no-escape < train.normalized.es > train.tokenized.es
    perl mosesdecoder/scripts/tokenizer/tokenizer.perl -l en -no-escape < train.normalized.en > train.tokenized.en
    

    For languages like Chinese where there is no explicit marker like spaces that separate words, we use Jieba to segment a string into words that are space separated.

    For example:

    Before - 同时,卫生局认为有必要接种的其他人员,包括公共部门,卫生局将主动联络有关机构取得名单后由卫生中心安排接种。
    After  - 同时 , 卫生局 认为 有 必要 接种 的 其他 人员 , 包括 公共部门 , 卫生局 将 主动 联络 有关 机构 取得 名单 后 由 卫生 中心 安排 接种 。
    

Training a BPE Tokenization#

Byte-pair encoding (BPE) [] is a sub-word tokenization algorithm that is commonly used to reduce the large vocabulary size of datasets by splitting words into frequently occuring sub-words. Currently, Machine translation only supports the YouTokenToMe BPE tokenizer. One can set the tokenization configuration as follows:

Parameter

Data Type

Default

Description

model.{encoder_tokenizer,decoder_tokenizer}.tokenizer_name

str

yttm

BPE library name. Only supports yttm for now.

model.{encoder_tokenizer,decoder_tokenizer}.tokenizer_model

str

null

Path to an existing YTTM BPE model. If null, will train one from scratch on the provided data.

model.{encoder_tokenizer,decoder_tokenizer}.vocab_size

int

null

Desired vocabulary size after BPE tokenization.

model.{encoder_tokenizer,decoder_tokenizer}.bpe_dropout

float

null

BPE dropout probability. [].

model.{encoder_tokenizer,decoder_tokenizer}.vocab_file

str

null

Path to pre-computed vocab file if exists.

model.shared_tokenizer

bool

True

Whether to share the tokenizer between the encoder and decoder.

Applying BPE Tokenization, Batching, Bucketing and Padding#

Given BPE tokenizers, and a cleaned parallel corpus, the following steps are applied to create a TranslationDataset object.

  1. Text to IDs - This performs subword tokenization with the BPE model on an input string and maps it to a sequence of tokens for the source and target text.

  2. Bucketing - Sentences vary in length and when creating minibatches, we’d like sentences in them to have roughly the same length to minimize the number of <pad> tokens and to maximize computational efficiency. This step groups sentences roughly the same length into buckets.

  3. Batching and padding - Creates minibatches with a maximum number of tokens specified by model.{train_ds,validation_ds,test_ds}.tokens_in_batch from buckets and pads, so they can be packed into a tensor.

Datasets can be configured as follows:

Parameter

Data Type

Default

Description

model.{train_ds,validation_ds,test_ds}.src_file_name

str

null

Path to the source language file.

model.{train_ds,validation_ds,test_ds}.tgt_file_name

str

null

Path to the target language file.

model.{train_ds,validation_ds,test_ds}.tokens_in_batch

int

512

Maximum number of tokens per minibatch.

model.{train_ds,validation_ds,test_ds}.clean

bool

true

Whether to clean the dataset by discarding examples that are greater than max_seq_length.

model.{train_ds,validation_ds,test_ds}.max_seq_length

int

512

Maximum sequence to be used with the clean argument above.

model.{train_ds,validation_ds,test_ds}.shuffle

bool

true

Whether to shuffle minibatches in the PyTorch DataLoader.

model.{train_ds,validation_ds,test_ds}.num_samples

int

-1

Number of samples to use. -1 for the entire dataset.

model.{train_ds,validation_ds,test_ds}.drop_last

bool

false

Drop last minibatch if it is not of equal size to the others.

model.{train_ds,validation_ds,test_ds}.pin_memory

bool

false

Whether to pin memory in the PyTorch DataLoader.

model.{train_ds,validation_ds,test_ds}.num_workers

int

8

Number of workers for the PyTorch DataLoader.

Tarred Datasets for Large Corpora#

When training with DistributedDataParallel, each process has its own copy of the dataset. For large datasets, this may not always fit in CPU memory. Webdatasets circumvents this problem by efficiently iterating over tar files stored on disk. Each tar file can contain hundreds to thousands of pickle files, each containing a single minibatch.

We recommend using this method when working with datasets with > 1 million sentence pairs.

Tarred datasets can be configured as follows:

Parameter

Data Type

Default

Description

model.{train_ds,validation_ds,test_ds}.use_tarred_dataset

bool

false

Whether to use tarred datasets.

model.{train_ds,validation_ds,test_ds}.tar_files

str

null

String specifying path to all tar files. Example with 100 tarfiles /path/to/tarfiles._OP_1..100_CL_.tar.

model.{train_ds,validation_ds,test_ds}.metadata_file

str

null

Path to JSON metadata file that contains only a single entry for the total number of batches in the dataset.

model.{train_ds,validation_ds,test_ds}.lines_per_dataset_fragment

int

1000000

Number of lines to consider for bucketing and padding.

model.{train_ds,validation_ds,test_ds}.num_batches_per_tarfile

int

100

Number of batches (pickle files) within each tarfile.

model.{train_ds,validation_ds,test_ds}.tar_shuffle_n

int

100

How many samples to look ahead and load to be shuffled.

model.{train_ds,validation_ds,test_ds}.shard_strategy

str

scatter

How the shards are distributed between multiple workers.

model.preproc_out_dir

str

null

Path to folder that contains processed tar files or directory where new tar files are written.

Tarred datasets can be created in two ways:

  1. Using the Hydra config and training script.

    For example:

    python examples/nlp/machine_translation/enc_dec_nmt.py \
      -cn aayn_base \
      do_training=false \
      model.preproc_out_dir=/path/to/preproc_dir \
      model.train_ds.use_tarred_dataset=true \
      model.train_ds.lines_per_dataset_fragment=1000000 \
      model.train_ds.num_batches_per_tarfile=200 \
      model.train_ds.src_file_name=train.tokenized.en \
      model.train_ds.tgt_file_name=train.tokenized.es \
      model.validation_ds.src_file_name=validation.tokenized.en \
      model.validation_ds.tgt_file_name=validation.tokenized.es \
      model.encoder_tokenizer.vocab_size=32000 \
      model.decoder_tokenizer.vocab_size=32000 \
      ~model.test_ds \
      trainer.devices=[0,1,2,3] \
      trainer.accelerator='gpu' \
      +trainer.fast_dev_run=true \
      exp_manager=null \
    

    The above script processes the parallel tokenized text files into tarred datasets that are written to /path/to/preproc_dir. Since do_training is set to False, the above script only creates tarred datasets and then exits. If do_training is set True, then one of two things happen:

    1. If no tar files are present in model.preproc_out_dir, the script first creates those files and then commences training.

    2. If tar files are already present in model.preproc_out_dir, the script starts training from the provided tar files.

  2. Using a separate script without Hydra.

    Tarred datasets for parallel corpora can also be created with a script that doesn’t require specifying a configs via Hydra and just uses Python argparse.

    For example:

    python examples/nlp/machine_translation/create_tarred_parallel_dataset.py \
      --shared_tokenizer \
      --clean \
      --bpe_dropout 0.1 \
      --src_fname train.tokenized.en \
      --tgt_fname train.tokenized.es \
      --out_dir /path/to/preproc_dir \
      --vocab_size 32000 \
      --max_seq_length 512 \
      --min_seq_length 1 \
      --tokens_in_batch 8192 \
      --lines_per_dataset_fragment 1000000 \
     --num_batches_per_tarfile 200
    

You can then set model.preproc_out_dir=/path/to/preproc_dir and model.train_ds.use_tarred_dataset=true to train with this data.

Model Configuration and Training#

The overall model consists of an encoder, decoder, and classification head. Encoders and decoders have the following configuration options:

Parameter

Data Type

Default

Description

model.{encoder,decoder}.max_sequence_length

int

512

Maximum sequence length of positional encodings.

model.{encoder,decoder}.embedding_dropout

float

0.1

Path to JSON metadata file that contains only a single entry for the total number of batches in the dataset.

model.{encoder,decoder}.learn_positional_encodings

bool

false

If True, this is a regular learnable embedding layer. If False, fixes position encodings to sinusoidal.

model.{encoder,decoder}.hidden_size

int

512

Size of the transformer hidden states.

model.{encoder,decoder}.num_layers

int

6

Number of transformer layers.

model.{encoder,decoder}.inner_size

int

2048

Size of the hidden states within the feedforward layers.

model.{encoder,decoder}.num_attention_heads

int

8

Number of attention heads.

model.{encoder,decoder}.ffn_dropout

float

0.1

Dropout probability within the feedforward layers.

model.{encoder,decoder}.attn_score_dropout

float

0.1

Dropout probability of the attention scores before softmax normalization.

model.{encoder,decoder}.attn_layer_dropout

float

0.1

Dropout probability of the attention query, key, and value projection activations.

model.{encoder,decoder}.hidden_act

str

relu

Activation function throughout the network.

model.{encoder,decoder}.mask_future

bool

false, true

Whether to mask future timesteps for attention. Defaults to True for decoder and False for encoder.

model.{encoder,decoder}.pre_ln

bool

false

Whether to apply layer-normalization before (true) or after (false) a sub-layer.

Our pre-trained models are optimized with Adam, with a maximum learning of 0.0004, beta of (0.9, 0.98), and inverse square root learning rate schedule from []. The model.optim section sets the optimization parameters.

The following script creates tarred datasets based on the provided parallel corpus and trains a model based on the base configuration from [].

python examples/nlp/machine_translation/enc_dec_nmt.py \
  -cn aayn_base \
  do_training=true \
  trainer.devices=8 \
  trainer.accelerator='gpu' \
  ~trainer.max_epochs \
  +trainer.max_steps=100000 \
  +trainer.val_check_interval=1000 \
  +exp_manager.exp_dir=/path/to/store/results \
  +exp_manager.create_checkpoint_callback=True \
  +exp_manager.checkpoint_callback_params.monitor=val_sacreBLEU \
  +exp_manager.checkpoint_callback_params.mode=max \
  +exp_manager.checkpoint_callback_params.save_top_k=5 \
  model.preproc_out_dir=/path/to/preproc_dir \
  model.train_ds.use_tarred_dataset=true \
  model.train_ds.lines_per_dataset_fragment=1000000 \
  model.train_ds.num_batches_per_tarfile=200 \
  model.train_ds.src_file_name=train.tokenized.en \
  model.train_ds.tgt_file_name=train.tokenized.es \
  model.validation_ds.src_file_name=validation.tokenized.en \
  model.validation_ds.tgt_file_name=validation.tokenized.es \
  model.encoder_tokenizer.vocab_size=32000 \
  model.decoder_tokenizer.vocab_size=32000 \
  ~model.test_ds \

The trainer keeps track of the sacreBLEU score [] on the provided validation set and saves the checkpoints that have the top 5 (by default) sacreBLEU scores.

At the end of training, a .nemo file is written to the result directory which allows to run inference on a test set.

Multi-Validation#

To run validation on multiple datasets, specify validation_ds.src_file_name and validation_ds.tgt_file_name with a list of file paths:

model.validation_ds.src_file_name=[/data/wmt13-en-de.src,/data/wmt14-en-de.src] \
model.validation_ds.tgt_file_name=[/data/wmt13-en-de.ref,/data/wmt14-en-de.ref] \

When using val_loss or val_sacreBLEU for the exp_manager.checkpoint_callback_params.monitor then the 0th indexed dataset will be used as the monitor.

To use other indexes, append the index:

exp_manager.checkpoint_callback_params.monitor=val_sacreBLEU_dl_index_1

Multiple test datasets work exactly the same way as validation datasets, simply replace validation_ds by test_ds in the above examples.

Bottleneck Models and Latent Variable Models (VAE, MIM)#

NMT with bottleneck encoder architecture is also supported (i.e., fixed size bottleneck), along with the training of Latent Variable Models (currently VAE, and MIM).

  1. Supported learning frameworks (model.model_type):
    • NLL - Conditional cross entropy (the usual NMT loss)

    • VAE - Variational Auto-Encoder (paper)

    • MIM - Mutual Information Machine (paper)

  2. Supported encoder architectures (model.encoder.arch):
    • seq2seq - the usual transformer encoder without a bottleneck

    • bridge - attention bridge bottleneck (paper)

    • perceiver - Perceiver bottleneck (paper)

Parameter

Data Type

Default

Description

model.model_type

str

nll

Learning (i.e., loss) type: nll (i.e., cross-entropy/auto-encoder), mim, vae (see description above)

model.min_logv

float

-6

Minimal allowed log variance for mim

model.latent_size

int

-1

Dimension of latent (projected from hidden) -1 will take value of hidden size

model. non_recon_warmup_batches

bool

200000

Warm-up steps for mim, and vae losses (anneals non-reconstruction part)

model. recon_per_token

bool

true

When false reconstruction is computed per sample, not per token

model.encoder.arch

str

seq2seq

Supported architectures: seq2seq, bridge, perceiver (see description above).

model.encoder.hidden_steps

int

32

Fixed number of hidden steps

model.encoder.hidden_blocks

int

1

Number of repeat blocks (see classes for description)

model.encoder. hidden_init_method

str

default

See classes for available values

Detailed description of config parameters:

  • model.encoder.arch=seq2seq
    • model.encoder.hidden_steps is ignored

    • model.encoder.hidden_blocks is ignored

    • model.encoder.hidden_init_method is ignored

  • model.encoder.arch=bridge
    • model.encoder.hidden_steps: input is projected to the specified fixed steps

    • model.encoder.hidden_blocks: number of encoder blocks to repeat after attention bridge projection

    • model.encoder.hidden_init_method:
      • enc_shared (default) - apply encoder to inputs, than attention bridge, followed by hidden_blocks number of the same encoder (pre and post encoders share parameters)

      • identity - apply attention bridge to inputs, followed by hidden_blocks number of the same encoder

      • enc - similar to enc_shared but the initial encoder has independent parameters

  • model.encoder.arch=perceiver
    • model.encoder.hidden_steps: input is projected to the specified fixed steps

    • model.encoder.hidden_blocks: number of cross-attention + self-attention blocks to repeat after initialization block (all self-attention and cross-attention share parameters)

    • model.encoder.hidden_init_method:
      • params (default) - hidden state is initialized with learned parameters followed by cross-attention with independent parameters

      • bridge - hidden state is initialized with an attention bridge

Training requires the use of the following script (instead of enc_dec_nmt.py):

python -- examples/nlp/machine_translation/enc_dec_nmt-bottleneck.py \
      --config-path=conf \
      --config-name=aayn_bottleneck \
      ...
      model.model_type=nll \
      model.non_recon_warmup_batches=7500 \
      model.encoder.arch=perceiver \
      model.encoder.hidden_steps=32 \
      model.encoder.hidden_blocks=2 \
      model.encoder.hidden_init_method=params \
      ...

Model Inference#

To generate translations on a test set and compute sacreBLEU scores, run the inference script:

python examples/nlp/machine_translation/nmt_transformer_infer.py \
  --model /path/to/model.nemo \
  --srctext test.en \
  --tgtout test.en-es.translations \
  --batch_size 128 \
  --source_lang en \
  --target_lang es

The --srctext file must be provided before tokenization and normalization. The resulting --tgtout file is detokenized and can be used to compute sacreBLEU scores.

cat test.en-es.translations | sacrebleu test.es

Inference Improvements#

In practice, there are a few commonly used techniques at inference to improve translation quality. NeMo implements:

  1. Model Ensembling

  2. Shallow Fusion decoding with transformer language models []

  3. Noisy-channel re-ranking []

  1. Model Ensembling - Given many models trained with the same encoder and decoder tokenizer, it is possible to ensemble their predictions (by averaging probabilities at each step) to generate better translations.

\[P(y_t|y_{<t},x;\theta_{1} \ldots \theta_{k}) = \frac{1}{k} \sum_{i=1}^k P(y_t|y_{<t},x;\theta_{i})\]

NOTE: It is important to make sure that all models being ensembled are trained with the same tokenizer.

The inference script will ensemble all models provided via the –model argument as a comma separated string pointing to multiple model paths.

For example, to ensemble three models /path/to/model1.nemo, /path/to/model2.nemo, /path/to/model3.nemo, run:

python examples/nlp/machine_translation/nmt_transformer_infer.py \
  --model /path/to/model1.nemo,/path/to/model2.nemo,/path/to/model3.nemo \
  --srctext test.en \
  --tgtout test.en-es.translations \
  --batch_size 128 \
  --source_lang en \
  --target_lang es
  1. Shallow Fusion Decoding with Transformer Language Models - Given a translation model or an ensemble ot translation models, it possible to combine the scores provided by the translation model(s) and a target-side language model.

At each decoding step, the score for a particular hypothesis on the beam is given by the weighted sum of the translation model log-probabilities and lanuage model log-probabilities.

\[\mathcal{S}(y_{1\ldots n}|x;\theta_{s \rightarrow t},\theta_{t}) = \mathcal{S}(y_{1\ldots n - 1}|x;\theta_{s \rightarrow t},\theta_{t}) + \log P(y_{n}|y_{<n},x;\theta_{s \rightarrow t}) + \lambda_{sf} \log P(y_{n}|y_{<n};\theta_{t})\]

Lambda controls the weight assigned to the language model. For now, the only family of language models supported are transformer language models trained in NeMo.

NOTE: The transformer language model needs to be trained using the same tokenizer as the decoder tokenizer in the NMT system.

For example, to ensemble three models /path/to/model1.nemo, /path/to/model2.nemo, /path/to/model3.nemo, with shallow fusion using an LM /path/to/lm.nemo

python examples/nlp/machine_translation/nmt_transformer_infer.py \
  --model /path/to/model1.nemo,/path/to/model2.nemo,/path/to/model3.nemo \
  --lm_model /path/to/lm.nemo \
  --fusion_coef 0.05 \
  --srctext test.en \
  --tgtout test.en-es.translations \
  --batch_size 128 \
  --source_lang en \
  --target_lang es
  1. Noisy Channel Re-ranking - Unlike ensembling and shallow fusion, noisy channel re-ranking only re-ranks the final candidates produced by beam search. It does so based on three scores

  1. Forward (source to target) translation model(s) log-proabilities

  2. Reverse (target to source) translation model(s) log-proabilities

  3. Language Model (target) log-proabilities

\[\argmax_{i} \mathcal{S}(y_i|x) = \log P(y_i|x;\theta_{s \rightarrow t}^{ens}) + \lambda_{ncr} \big( \log P(x|y_i;\theta_{t \rightarrow s}) + \log P(y_i;\theta_{t}) \big)\]

To perform noisy-channel re-ranking, first generate a .scores file that contains log-proabilities from the forward translation model for each hypothesis on the beam.

This will generate a scores file test.en-es.translations.scores, which is provided as input to NeMo/examples/nlp/machine_translation/noisy_channel_reranking.py

This script also requires a reverse (target to source) translation model and a target language model.

python noisy_channel_reranking.py \
    --reverse_model=/path/to/reverse_model1.nemo,/path/to/reverse_model2.nemo \
    --language_model=/path/to/lm.nemo \
    --srctext=test.en-es.translations.scores \
    --tgtout=test-en-es.ncr.translations \
    --forward_model_coef=1.0 \
    --reverse_model_coef=0.7 \
    --target_lm_coef=0.05 \

Pretrained Encoders#

Pretrained BERT encoders from either HuggingFace Transformers or Megatron-LM can be used to to train NeMo NMT models.

The library flag takes values: huggingface, megatron, and nemo.

The model_name flag is used to indicate a named model architecture. For example, we can use bert_base_cased from HuggingFace or megatron-bert-345m-cased from Megatron-LM.

The pretrained flag indicates whether or not to download the pretrained weights (pretrained=True) or instantiate the same model architecture with random weights (pretrained=False).

To use a custom model architecture from a specific library, use model_name=null and then add the custom configuration under the encoder configuration.

HuggingFace#

We have provided a HuggingFace config file to use with HuggingFace encoders.

To use the config file from CLI:

--config-path=conf \
--config-name=huggingface \

As an example, we can configure the NeMo NMT encoder to use bert-base-cased from HuggingFace by using the huggingface config file and setting

model.encoder.pretrained=true \
model.encoder.model_name=bert-base-cased \

To use a custom architecture from HuggingFace we can use

+model.encoder._target_=transformers.BertConfig \
+model.encoder.hidden_size=1536 \

Note the + symbol is needed if we’re not adding the arguments to the YAML config file.

Megatron#

We have provided a Megatron config file to use with Megatron encoders.

To use the config file from CLI:

--config-path=conf \
--config-name=megatron \

The checkpoint_file should be the path to Megatron-LM checkpoint:

/path/to/your/megatron/checkpoint/model_optim_rng.pt

In case your megatron model requires model parallelism, then checkpoint_file should point to the directory containing the standard Megatron-LM checkpoint format:

3.9b_bert_no_rng
├── mp_rank_00
│   └── model_optim_rng.pt
├── mp_rank_01
│   └── model_optim_rng.pt
├── mp_rank_02
│   └── model_optim_rng.pt
└── mp_rank_03
    └── model_optim_rng.pt

As an example, to train a NeMo NMT model with a 3.9B Megatron BERT encoder, we would use the following encoder configuration:

model.encoder.checkpoint_file=/path/to/megatron/checkpoint/3.9b_bert_no_rng \
model.encoder.hidden_size=2560 \
model.encoder.num_attention_heads=40 \
model.encoder.num_layers=48 \
model.encoder.max_position_embeddings=512 \

To train a Megatron 345M BERT, we would use

model.encoder.model_name=megatron-bert-cased \
model.encoder.checkpoint_file=/path/to/your/megatron/checkpoint/model_optim_rng.pt \
model.encoder.hidden_size=1024 \
model.encoder.num_attention_heads=16 \
model.encoder.num_layers=24 \
model.encoder.max_position_embeddings=512 \

If the pretrained megatron model used a custom vocab file, then set:

model.encoder_tokenizer.vocab_file=/path/to/your/megatron/vocab_file.txt
model.encoder.vocab_file=/path/to/your/megatron/vocab_file.txt

Use encoder.model_name=megatron_bert_uncased for uncased models with custom vocabularies and use encoder.model_name=megatron_bert_cased for cased models with custom vocabularies.

References#