Important
NeMo 2.0 is an experimental feature and currently released in the dev container only: nvcr.io/nvidia/nemo:dev. Please refer to NeMo 2.0 overview for information on getting started.
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 [nlp-machine_translation5].
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
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.
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 [nlp-machine_translation4].
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.
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
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.binmodel 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
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
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
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
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
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.
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
NeMoandNeMo.. Tokenization splits punctuation from the word to create two separate tokens. In the previous exampleNeMo.becomesNeMo .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) [nlp-machine_translation4] 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 |
|
BPE library name. Only supports |
model.{encoder_tokenizer,decoder_tokenizer}.tokenizer_model |
str |
|
Path to an existing YTTM BPE model. If |
model.{encoder_tokenizer,decoder_tokenizer}.vocab_size |
int |
|
Desired vocabulary size after BPE tokenization. |
model.{encoder_tokenizer,decoder_tokenizer}.bpe_dropout |
float |
|
BPE dropout probability. [nlp-machine_translation3]. |
model.{encoder_tokenizer,decoder_tokenizer}.vocab_file |
str |
|
Path to pre-computed vocab file if exists. |
model.shared_tokenizer |
bool |
|
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.
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.
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.Batching and padding - Creates minibatches with a maximum number of tokens specified by
model.{train_ds,validation_ds,test_ds}.tokens_in_batchfrom 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 |
|
Path to the source language file. |
model.{train_ds,validation_ds,test_ds}.tgt_file_name |
str |
|
Path to the target language file. |
model.{train_ds,validation_ds,test_ds}.tokens_in_batch |
int |
|
Maximum number of tokens per minibatch. |
model.{train_ds,validation_ds,test_ds}.clean |
bool |
|
Whether to clean the dataset by discarding examples that are greater than |
model.{train_ds,validation_ds,test_ds}.max_seq_length |
int |
|
Maximum sequence to be used with the |
model.{train_ds,validation_ds,test_ds}.shuffle |
bool |
|
Whether to shuffle minibatches in the PyTorch DataLoader. |
model.{train_ds,validation_ds,test_ds}.num_samples |
int |
|
Number of samples to use. |
model.{train_ds,validation_ds,test_ds}.drop_last |
bool |
|
Drop last minibatch if it is not of equal size to the others. |
model.{train_ds,validation_ds,test_ds}.pin_memory |
bool |
|
Whether to pin memory in the PyTorch DataLoader. |
model.{train_ds,validation_ds,test_ds}.num_workers |
int |
|
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 |
|
Whether to use tarred datasets. |
model.{train_ds,validation_ds,test_ds}.tar_files |
str |
|
String specifying path to all tar files. Example with 100 tarfiles |
model.{train_ds,validation_ds,test_ds}.metadata_file |
str |
|
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 |
|
Number of lines to consider for bucketing and padding. |
model.{train_ds,validation_ds,test_ds}.num_batches_per_tarfile |
int |
|
Number of batches (pickle files) within each tarfile. |
model.{train_ds,validation_ds,test_ds}.tar_shuffle_n |
int |
|
How many samples to look ahead and load to be shuffled. |
model.{train_ds,validation_ds,test_ds}.shard_strategy |
str |
|
How the shards are distributed between multiple workers. |
model.preproc_out_dir |
str |
|
Path to folder that contains processed tar files or directory where new tar files are written. |
Tarred datasets can be created in two ways:
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. Sincedo_trainingis set toFalse, the above script only creates tarred datasets and then exits. Ifdo_trainingis setTrue, then one of two things happen:If no tar files are present in
model.preproc_out_dir, the script first creates those files and then commences training.If tar files are already present in
model.preproc_out_dir, the script starts training from the provided tar files.
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 |
|
Maximum sequence length of positional encodings. |
model.{encoder,decoder}.embedding_dropout |
float |
|
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 |
|
If |
model.{encoder,decoder}.hidden_size |
int |
|
Size of the transformer hidden states. |
model.{encoder,decoder}.num_layers |
int |
|
Number of transformer layers. |
model.{encoder,decoder}.inner_size |
int |
|
Size of the hidden states within the feedforward layers. |
model.{encoder,decoder}.num_attention_heads |
int |
|
Number of attention heads. |
model.{encoder,decoder}.ffn_dropout |
float |
|
Dropout probability within the feedforward layers. |
model.{encoder,decoder}.attn_score_dropout |
float |
|
Dropout probability of the attention scores before softmax normalization. |
model.{encoder,decoder}.attn_layer_dropout |
float |
|
Dropout probability of the attention query, key, and value projection activations. |
model.{encoder,decoder}.hidden_act |
str |
|
Activation function throughout the network. |
model.{encoder,decoder}.mask_future |
bool |
|
Whether to mask future timesteps for attention. Defaults to |
model.{encoder,decoder}.pre_ln |
bool |
|
Whether to apply layer-normalization before ( |
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 [nlp-machine_translation5]. 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 [nlp-machine_translation5].
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 [nlp-machine_translation2] 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).
Parameter |
Data Type |
Default |
Description |
|---|---|---|---|
model.model_type |
str |
|
Learning (i.e., loss) type: nll (i.e., cross-entropy/auto-encoder), mim, vae (see description above) |
model.min_logv |
float |
|
Minimal allowed log variance for mim |
model.latent_size |
int |
|
Dimension of latent (projected from hidden) -1 will take value of hidden size |
model. non_recon_warmup_batches |
bool |
|
Warm-up steps for mim, and vae losses (anneals non-reconstruction part) |
model. recon_per_token |
bool |
|
When false reconstruction is computed per sample, not per token |
model.encoder.arch |
str |
|
Supported architectures: |
model.encoder.hidden_steps |
int |
|
Fixed number of hidden steps |
model.encoder.hidden_blocks |
int |
|
Number of repeat blocks (see classes for description) |
model.encoder. hidden_init_method |
str |
|
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:
Model Ensembling
Shallow Fusion decoding with transformer language models [nlp-machine_translation1]
Noisy-channel re-ranking [nlp-machine_translation6]
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.
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
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.
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
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
Forward (source to target) translation model(s) log-probabilities
Reverse (target to source) translation model(s) log-probabilities
Language Model (target) log-probabilities
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.
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 \
--write_scores \
--fusion_coef 0.05 \
--srctext test.en \
--tgtout test.en-es.translations \
--batch_size 128 \
--source_lang en \
--target_lang es
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
- nlp-machine_translation1
Caglar Gulcehre, Orhan Firat, Kelvin Xu, Kyunghyun Cho, Loic Barrault, Huei-Chi Lin, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. On using monolingual corpora in neural machine translation. arXiv preprint arXiv:1503.03535, 2015.
- nlp-machine_translation2
Matt Post. A call for clarity in reporting bleu scores. arXiv preprint arXiv:1804.08771, 2018.
- nlp-machine_translation3
Ivan Provilkov, Dmitrii Emelianenko, and Elena Voita. Bpe-dropout: simple and effective subword regularization. arXiv preprint arXiv:1910.13267, 2019.
- nlp-machine_translation4(1,2)
Rico Sennrich, Barry Haddow, and Alexandra Birch. Neural machine translation of rare words with subword units. arXiv preprint arXiv:1508.07909, 2015.
- nlp-machine_translation5(1,2,3)
Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. In Advances in Neural Information Processing Systems, 6000–6010. 2017.
- nlp-machine_translation6
Kyra Yee, Nathan Ng, Yann N Dauphin, and Michael Auli. Simple and effective noisy channel modeling for neural machine translation. arXiv preprint arXiv:1908.05731, 2019.