class MTEncDecModel(EncDecNLPModel):
"""
Encoder-decoder machine translation model.
"""
def __init__(self, cfg: MTEncDecModelConfig, trainer: Trainer = None):
cfg = model_utils.convert_model_config_to_dict_config(cfg)
# Get global rank and total number of GPU workers for IterableDataset partitioning, if applicable
# Global_rank and local_rank is set by LightningModule in Lightning 1.2.0
self.world_size = 1
if trainer is not None:
self.world_size = trainer.num_nodes * trainer.num_gpus
cfg = model_utils.maybe_update_config_version(cfg)
self.src_language = cfg.get("src_language", None)
self.tgt_language = cfg.get("tgt_language", None)
self.multilingual = cfg.get("multilingual", False)
self.multilingual_ids = []
self.encoder_tokenizer_library = cfg.encoder_tokenizer.get(
'library', 'yttm')
self.decoder_tokenizer_library = cfg.decoder_tokenizer.get(
'library', 'yttm')
# Instantiates tokenizers and register to be saved with NeMo Model archive
# After this call, ther will be self.encoder_tokenizer and self.decoder_tokenizer
# Which can convert between tokens and token_ids for SRC and TGT languages correspondingly.
self.setup_enc_dec_tokenizers(
encoder_tokenizer_library=self.encoder_tokenizer_library,
encoder_tokenizer_model=cfg.encoder_tokenizer.get(
'tokenizer_model'),
encoder_bpe_dropout=cfg.encoder_tokenizer.get(
'bpe_dropout', 0.0) if cfg.encoder_tokenizer.get(
'bpe_dropout', 0.0) is not None else 0.0,
encoder_model_name=cfg.encoder.get('model_name') if hasattr(
cfg.encoder, 'model_name') else None,
encoder_r2l=cfg.encoder_tokenizer.get('r2l', False),
decoder_tokenizer_library=self.decoder_tokenizer_library,
encoder_tokenizer_vocab_file=cfg.encoder_tokenizer.get(
'vocab_file', None),
decoder_tokenizer_model=cfg.decoder_tokenizer.tokenizer_model,
decoder_bpe_dropout=cfg.decoder_tokenizer.get(
'bpe_dropout', 0.0) if cfg.decoder_tokenizer.get(
'bpe_dropout', 0.0) is not None else 0.0,
decoder_model_name=cfg.decoder.get('model_name') if hasattr(
cfg.decoder, 'model_name') else None,
decoder_r2l=cfg.decoder_tokenizer.get('r2l', False),
)
if self.multilingual:
if isinstance(self.src_language, ListConfig) and isinstance(
self.tgt_language, ListConfig):
raise ValueError(
"cfg.src_language and cfg.tgt_language cannot both be lists. We only support many-to-one or one-to-many multilingual models."
)
elif isinstance(self.src_language, ListConfig):
for lng in self.src_language:
self.multilingual_ids.append(
self.encoder_tokenizer.token_to_id("<" + lng + ">"))
elif isinstance(self.tgt_language, ListConfig):
for lng in self.tgt_language:
self.multilingual_ids.append(
self.encoder_tokenizer.token_to_id("<" + lng + ">"))
else:
raise ValueError(
"Expect either cfg.src_language or cfg.tgt_language to be a list when multilingual=True."
)
if isinstance(self.src_language, ListConfig):
self.tgt_language = [self.tgt_language] * len(
self.src_language)
else:
self.src_language = [self.src_language] * len(
self.tgt_language)
self.source_processor_list = []
self.target_processor_list = []
for src_lng, tgt_lng in zip(self.src_language, self.tgt_language):
src_prcsr, tgt_prscr = self.setup_pre_and_post_processing_utils(
src_lng, tgt_lng)
self.source_processor_list.append(src_prcsr)
self.target_processor_list.append(tgt_prscr)
else:
# After this call, the model will have self.source_processor and self.target_processor objects
self.setup_pre_and_post_processing_utils(self.src_language,
self.tgt_language)
self.multilingual_ids = [None]
# TODO: Why is this base constructor call so late in the game?
super().__init__(cfg=cfg, trainer=trainer)
# encoder from NeMo, Megatron-LM, or HuggingFace
encoder_cfg_dict = OmegaConf.to_container(cfg.get('encoder'))
encoder_cfg_dict['vocab_size'] = self.encoder_vocab_size
library = encoder_cfg_dict.pop('library', 'nemo')
model_name = encoder_cfg_dict.pop('model_name', None)
pretrained = encoder_cfg_dict.pop('pretrained', False)
checkpoint_file = encoder_cfg_dict.pop('checkpoint_file', None)
self.encoder = get_transformer(
library=library,
model_name=model_name,
pretrained=pretrained,
config_dict=encoder_cfg_dict,
encoder=True,
pre_ln_final_layer_norm=encoder_cfg_dict.get(
'pre_ln_final_layer_norm', False),
checkpoint_file=checkpoint_file,
)
# decoder from NeMo, Megatron-LM, or HuggingFace
decoder_cfg_dict = OmegaConf.to_container(cfg.get('decoder'))
decoder_cfg_dict['vocab_size'] = self.decoder_vocab_size
library = decoder_cfg_dict.pop('library', 'nemo')
model_name = decoder_cfg_dict.pop('model_name', None)
pretrained = decoder_cfg_dict.pop('pretrained', False)
decoder_cfg_dict['hidden_size'] = self.encoder.hidden_size
self.decoder = get_transformer(
library=library,
model_name=model_name,
pretrained=pretrained,
config_dict=decoder_cfg_dict,
encoder=False,
pre_ln_final_layer_norm=decoder_cfg_dict.get(
'pre_ln_final_layer_norm', False),
)
self.log_softmax = TokenClassifier(
hidden_size=self.decoder.hidden_size,
num_classes=self.decoder_vocab_size,
activation=cfg.head.activation,
log_softmax=cfg.head.log_softmax,
dropout=cfg.head.dropout,
use_transformer_init=cfg.head.use_transformer_init,
)
self.beam_search = BeamSearchSequenceGenerator(
embedding=self.decoder.embedding,
decoder=self.decoder.decoder,
log_softmax=self.log_softmax,
max_sequence_length=self.decoder.max_sequence_length,
beam_size=cfg.beam_size,
bos=self.decoder_tokenizer.bos_id,
pad=self.decoder_tokenizer.pad_id,
eos=self.decoder_tokenizer.eos_id,
len_pen=cfg.len_pen,
max_delta_length=cfg.max_generation_delta,
)
# tie weights of embedding and softmax matrices
self.log_softmax.mlp.layer0.weight = self.decoder.embedding.token_embedding.weight
# TODO: encoder and decoder with different hidden size?
std_init_range = 1 / self.encoder.hidden_size**0.5
# initialize weights if not using pretrained encoder/decoder
if not self._cfg.encoder.get('pretrained', False):
self.encoder.apply(lambda module: transformer_weights_init(
module, std_init_range))
if not self._cfg.decoder.get('pretrained', False):
self.decoder.apply(lambda module: transformer_weights_init(
module, std_init_range))
self.log_softmax.apply(
lambda module: transformer_weights_init(module, std_init_range))
self.loss_fn = SmoothedCrossEntropyLoss(
pad_id=self.decoder_tokenizer.pad_id,
label_smoothing=cfg.label_smoothing)
self.eval_loss_fn = NLLLoss(ignore_index=self.decoder_tokenizer.pad_id)
def filter_predicted_ids(self, ids):
ids[ids >=
self.decoder_tokenizer.vocab_size] = self.decoder_tokenizer.unk_id
return ids
@typecheck()
def forward(self, src, src_mask, tgt, tgt_mask):
src_hiddens = self.encoder(input_ids=src, encoder_mask=src_mask)
tgt_hiddens = self.decoder(input_ids=tgt,
decoder_mask=tgt_mask,
encoder_embeddings=src_hiddens,
encoder_mask=src_mask)
log_probs = self.log_softmax(hidden_states=tgt_hiddens)
return log_probs
def training_step(self, batch, batch_idx):
"""
Lightning calls this inside the training loop with the data from the training dataloader
passed in as `batch`.
"""
# forward pass
for i in range(len(batch)):
if batch[i].ndim == 3:
# Dataset returns already batched data and the first dimension of size 1 added by DataLoader
# is excess.
batch[i] = batch[i].squeeze(dim=0)
src_ids, src_mask, tgt_ids, tgt_mask, labels = batch
log_probs = self(src_ids, src_mask, tgt_ids, tgt_mask)
train_loss = self.loss_fn(log_probs=log_probs, labels=labels)
tensorboard_logs = {
'train_loss': train_loss,
'lr': self._optimizer.param_groups[0]['lr'],
}
return {'loss': train_loss, 'log': tensorboard_logs}
def eval_step(self, batch, batch_idx, mode, dataloader_idx=0):
for i in range(len(batch)):
if batch[i].ndim == 3:
# Dataset returns already batched data and the first dimension of size 1 added by DataLoader
# is excess.
batch[i] = batch[i].squeeze(dim=0)
if self.multilingual:
self.source_processor = self.source_processor_list[dataloader_idx]
self.target_processor = self.target_processor_list[dataloader_idx]
src_ids, src_mask, tgt_ids, tgt_mask, labels = batch
log_probs = self(src_ids, src_mask, tgt_ids, tgt_mask)
eval_loss = self.eval_loss_fn(log_probs=log_probs, labels=labels)
# this will run encoder twice -- TODO: potentially fix
_, translations = self.batch_translate(src=src_ids, src_mask=src_mask)
if dataloader_idx == 0:
getattr(self, f'{mode}_loss')(loss=eval_loss,
num_measurements=log_probs.shape[0] *
log_probs.shape[1])
else:
getattr(self, f'{mode}_loss_{dataloader_idx}')(
loss=eval_loss,
num_measurements=log_probs.shape[0] * log_probs.shape[1])
np_tgt = tgt_ids.detach().cpu().numpy()
ground_truths = [
self.decoder_tokenizer.ids_to_text(tgt) for tgt in np_tgt
]
ground_truths = [
self.target_processor.detokenize(tgt.split(' '))
for tgt in ground_truths
]
num_non_pad_tokens = np.not_equal(
np_tgt, self.decoder_tokenizer.pad_id).sum().item()
return {
'translations': translations,
'ground_truths': ground_truths,
'num_non_pad_tokens': num_non_pad_tokens,
}
def test_step(self, batch, batch_idx, dataloader_idx=0):
return self.eval_step(batch, batch_idx, 'test', dataloader_idx)
@rank_zero_only
def log_param_stats(self):
for name, p in self.named_parameters():
if p.requires_grad:
self.trainer.logger.experiment.add_histogram(
name + '_hist', p, global_step=self.global_step)
self.trainer.logger.experiment.add_scalars(
name,
{
'mean': p.mean(),
'stddev': p.std(),
'max': p.max(),
'min': p.min()
},
global_step=self.global_step,
)
def validation_step(self, batch, batch_idx, dataloader_idx=0):
"""
Lightning calls this inside the validation loop with the data from the validation dataloader
passed in as `batch`.
"""
return self.eval_step(batch, batch_idx, 'val', dataloader_idx)
def eval_epoch_end(self, outputs, mode):
# if user specifies one validation dataloader, then PTL reverts to giving a list of dictionary instead of a list of list of dictionary
if isinstance(outputs[0], dict):
outputs = [outputs]
loss_list = []
sb_score_list = []
for dataloader_idx, output in enumerate(outputs):
if dataloader_idx == 0:
eval_loss = getattr(self, f'{mode}_loss').compute()
else:
eval_loss = getattr(self,
f'{mode}_loss_{dataloader_idx}').compute()
translations = list(
itertools.chain(*[x['translations'] for x in output]))
ground_truths = list(
itertools.chain(*[x['ground_truths'] for x in output]))
assert len(translations) == len(ground_truths)
# Gather translations and ground truths from all workers
tr_and_gt = [None for _ in range(self.world_size)]
# we also need to drop pairs where ground truth is an empty string
dist.all_gather_object(
tr_and_gt, [(t, g)
for (t, g) in zip(translations, ground_truths)
if g.strip() != ''])
if self.global_rank == 0:
_translations = []
_ground_truths = []
for rank in range(0, self.world_size):
_translations += [t for (t, g) in tr_and_gt[rank]]
_ground_truths += [g for (t, g) in tr_and_gt[rank]]
if self.tgt_language in ['ja']:
sacre_bleu = corpus_bleu(_translations, [_ground_truths],
tokenize="ja-mecab")
elif self.tgt_language in ['zh']:
sacre_bleu = corpus_bleu(_translations, [_ground_truths],
tokenize="zh")
else:
sacre_bleu = corpus_bleu(_translations, [_ground_truths],
tokenize="13a")
# because the reduction op later is average (over word_size)
sb_score = sacre_bleu.score * self.world_size
dataset_name = "Validation" if mode == 'val' else "Test"
logging.info(
f"Dataset name: {dataset_name}, Dataloader index: {dataloader_idx}, Set size: {len(translations)}"
)
logging.info(
f"Dataset name: {dataset_name}, Dataloader index: {dataloader_idx}, Val Loss = {eval_loss}"
)
logging.info(
f"Dataset name: {dataset_name}, Dataloader index: {dataloader_idx}, Sacre BLEU = {sb_score / self.world_size}"
)
logging.info(
f"Dataset name: {dataset_name}, Dataloader index: {dataloader_idx}, Translation Examples:"
)
for i in range(0, 3):
ind = random.randint(0, len(translations) - 1)
logging.info(" " + '\u0332'.join(f"Example {i}:"))
logging.info(f" Prediction: {translations[ind]}")
logging.info(f" Ground Truth: {ground_truths[ind]}")
else:
sb_score = 0.0
loss_list.append(eval_loss.cpu().numpy())
sb_score_list.append(sb_score)
if dataloader_idx == 0:
self.log(f"{mode}_loss", eval_loss, sync_dist=True)
self.log(f"{mode}_sacreBLEU", sb_score, sync_dist=True)
getattr(self, f'{mode}_loss').reset()
else:
self.log(f"{mode}_loss_dl_index_{dataloader_idx}",
eval_loss,
sync_dist=True)
self.log(f"{mode}_sacreBLEU_dl_index_{dataloader_idx}",
sb_score,
sync_dist=True)
getattr(self, f'{mode}_loss_{dataloader_idx}').reset()
if len(loss_list) > 1:
self.log(f"{mode}_loss_avg", np.mean(loss_list), sync_dist=True)
self.log(f"{mode}_sacreBLEU_avg",
np.mean(sb_score_list),
sync_dist=True)
def validation_epoch_end(self, outputs):
"""
Called at the end of validation to aggregate outputs.
:param outputs: list of individual outputs of each validation step.
"""
self.eval_epoch_end(outputs, 'val')
def test_epoch_end(self, outputs):
self.eval_epoch_end(outputs, 'test')
def setup_enc_dec_tokenizers(
self,
encoder_tokenizer_library=None,
encoder_tokenizer_model=None,
encoder_bpe_dropout=0.0,
encoder_model_name=None,
encoder_r2l=False,
encoder_tokenizer_vocab_file=None,
decoder_tokenizer_library=None,
decoder_tokenizer_model=None,
decoder_bpe_dropout=0.0,
decoder_model_name=None,
decoder_r2l=False,
):
supported_tokenizers = [
'yttm', 'huggingface', 'sentencepiece', 'megatron', 'byte-level'
]
if (encoder_tokenizer_library not in supported_tokenizers
or decoder_tokenizer_library not in supported_tokenizers):
raise NotImplementedError(
f"Currently we only support tokenizers in {supported_tokenizers}."
)
self.encoder_tokenizer = get_nmt_tokenizer(
library=encoder_tokenizer_library,
tokenizer_model=self.register_artifact(
"encoder_tokenizer.tokenizer_model", encoder_tokenizer_model),
bpe_dropout=encoder_bpe_dropout,
model_name=encoder_model_name,
vocab_file=self.register_artifact("encoder_tokenizer.vocab_file",
encoder_tokenizer_vocab_file),
special_tokens=None,
use_fast=False,
r2l=encoder_r2l,
)
self.decoder_tokenizer = get_nmt_tokenizer(
library=decoder_tokenizer_library,
tokenizer_model=self.register_artifact(
"decoder_tokenizer.tokenizer_model", decoder_tokenizer_model),
bpe_dropout=decoder_bpe_dropout,
model_name=decoder_model_name,
vocab_file=None,
special_tokens=None,
use_fast=False,
r2l=decoder_r2l,
)
def setup_training_data(self, train_data_config: Optional[DictConfig]):
self._train_dl = self._setup_dataloader_from_config(
cfg=train_data_config)
def setup_multiple_validation_data(self, val_data_config: Union[DictConfig,
Dict]):
self.setup_validation_data(self._cfg.get('validation_ds'))
def setup_multiple_test_data(self, test_data_config: Union[DictConfig,
Dict]):
self.setup_test_data(self._cfg.get('test_ds'))
def setup_validation_data(self, val_data_config: Optional[DictConfig]):
self._validation_dl = self._setup_eval_dataloader_from_config(
cfg=val_data_config)
# instantiate Torchmetric for each val dataloader
if self._validation_dl is not None:
for dataloader_idx in range(len(self._validation_dl)):
if dataloader_idx == 0:
setattr(
self,
f'val_loss',
GlobalAverageLossMetric(dist_sync_on_step=False,
take_avg_loss=True),
)
else:
setattr(
self,
f'val_loss_{dataloader_idx}',
GlobalAverageLossMetric(dist_sync_on_step=False,
take_avg_loss=True),
)
def setup_test_data(self, test_data_config: Optional[DictConfig]):
self._test_dl = self._setup_eval_dataloader_from_config(
cfg=test_data_config)
# instantiate Torchmetric for each test dataloader
if self._test_dl is not None:
for dataloader_idx in range(len(self._test_dl)):
if dataloader_idx == 0:
setattr(
self,
f'test_loss',
GlobalAverageLossMetric(dist_sync_on_step=False,
take_avg_loss=True),
)
else:
setattr(
self,
f'test_loss_{dataloader_idx}',
GlobalAverageLossMetric(dist_sync_on_step=False,
take_avg_loss=True),
)
def _setup_dataloader_from_config(self, cfg: DictConfig):
if cfg.get("use_tarred_dataset", False):
if cfg.get("metadata_file") is None:
raise FileNotFoundError(
"Trying to use tarred data set but could not find metadata path in config."
)
metadata_file_list = cfg.get('metadata_file')
tar_files_list = cfg.get('tar_files', None)
if isinstance(metadata_file_list, str):
metadata_file_list = [metadata_file_list]
if tar_files_list is not None and isinstance(tar_files_list, str):
tar_files_list = [tar_files_list]
if tar_files_list is not None and len(tar_files_list) != len(
metadata_file_list):
raise ValueError(
'The config must have the same number of tarfile paths and metadata file paths.'
)
datasets = []
for idx, metadata_file in enumerate(metadata_file_list):
with open(metadata_file) as metadata_reader:
metadata = json.load(metadata_reader)
if tar_files_list is None:
tar_files = metadata.get('tar_files')
if tar_files is not None:
logging.info(
f'Loading from tarred dataset {tar_files}')
else:
tar_files = tar_files_list[idx]
if metadata.get('tar_files') is not None:
logging.info(
f'Tar file paths found in both cfg and metadata using one in cfg by default - {tar_files}'
)
dataset = TarredTranslationDataset(
text_tar_filepaths=tar_files,
metadata_path=metadata_file,
encoder_tokenizer=self.encoder_tokenizer,
decoder_tokenizer=self.decoder_tokenizer,
shuffle_n=cfg.get("tar_shuffle_n", 100),
shard_strategy=cfg.get("shard_strategy", "scatter"),
global_rank=self.global_rank,
world_size=self.world_size,
reverse_lang_direction=cfg.get("reverse_lang_direction",
False),
prepend_id=self.multilingual_ids[idx]
if self.multilingual else None,
)
datasets.append(dataset)
if len(datasets) > 1:
dataset = ConcatDataset(
datasets=datasets,
sampling_technique=cfg.get('concat_sampling_technique'),
sampling_temperature=cfg.get(
'concat_sampling_temperature'),
sampling_probabilities=cfg.get(
'concat_sampling_probabilities'),
global_rank=self.global_rank,
world_size=self.world_size,
)
else:
dataset = datasets[0]
else:
src_file_list = cfg.src_file_name
tgt_file_list = cfg.tgt_file_name
if isinstance(src_file_list, str):
src_file_list = [src_file_list]
if isinstance(tgt_file_list, str):
tgt_file_list = [tgt_file_list]
if len(src_file_list) != len(tgt_file_list):
raise ValueError(
'The same number of filepaths must be passed in for source and target.'
)
datasets = []
for idx, src_file in enumerate(src_file_list):
dataset = TranslationDataset(
dataset_src=str(Path(src_file).expanduser()),
dataset_tgt=str(Path(tgt_file_list[idx]).expanduser()),
tokens_in_batch=cfg.tokens_in_batch,
clean=cfg.get("clean", False),
max_seq_length=cfg.get("max_seq_length", 512),
min_seq_length=cfg.get("min_seq_length", 1),
max_seq_length_diff=cfg.get("max_seq_length_diff", 512),
max_seq_length_ratio=cfg.get("max_seq_length_ratio", 512),
cache_ids=cfg.get("cache_ids", False),
cache_data_per_node=cfg.get("cache_data_per_node", False),
use_cache=cfg.get("use_cache", False),
reverse_lang_direction=cfg.get("reverse_lang_direction",
False),
prepend_id=self.multilingual_ids[idx]
if self.multilingual else None,
)
dataset.batchify(self.encoder_tokenizer,
self.decoder_tokenizer)
datasets.append(dataset)
if len(datasets) > 1:
dataset = ConcatDataset(
datasets=datasets,
shuffle=cfg.get('shuffle'),
sampling_technique=cfg.get('concat_sampling_technique'),
sampling_temperature=cfg.get(
'concat_sampling_temperature'),
sampling_probabilities=cfg.get(
'concat_sampling_probabilities'),
global_rank=self.global_rank,
world_size=self.world_size,
)
else:
dataset = datasets[0]
if cfg.shuffle:
sampler = pt_data.RandomSampler(dataset)
else:
sampler = pt_data.SequentialSampler(dataset)
return torch.utils.data.DataLoader(
dataset=dataset,
batch_size=1,
sampler=None if cfg.get("use_tarred_dataset", False) else sampler,
num_workers=cfg.get("num_workers", 2),
pin_memory=cfg.get("pin_memory", False),
drop_last=cfg.get("drop_last", False),
)
def replace_beam_with_sampling(self, topk=500):
self.beam_search = TopKSequenceGenerator(
embedding=self.decoder.embedding,
decoder=self.decoder.decoder,
log_softmax=self.log_softmax,
max_sequence_length=self.beam_search.max_seq_length,
beam_size=topk,
bos=self.decoder_tokenizer.bos_id,
pad=self.decoder_tokenizer.pad_id,
eos=self.decoder_tokenizer.eos_id,
)
def _setup_eval_dataloader_from_config(self, cfg: DictConfig):
src_file_name = cfg.get('src_file_name')
tgt_file_name = cfg.get('tgt_file_name')
if src_file_name is None or tgt_file_name is None:
raise ValueError(
'Validation dataloader needs both cfg.src_file_name and cfg.tgt_file_name to not be None.'
)
else:
# convert src_file_name and tgt_file_name to list of strings
if isinstance(src_file_name, str):
src_file_list = [src_file_name]
elif isinstance(src_file_name, ListConfig):
src_file_list = src_file_name
else:
raise ValueError(
"cfg.src_file_name must be string or list of strings")
if isinstance(tgt_file_name, str):
tgt_file_list = [tgt_file_name]
elif isinstance(tgt_file_name, ListConfig):
tgt_file_list = tgt_file_name
else:
raise ValueError(
"cfg.tgt_file_name must be string or list of strings")
if len(src_file_list) != len(tgt_file_list):
raise ValueError(
'The same number of filepaths must be passed in for source and target validation.'
)
dataloaders = []
prepend_idx = 0
for idx, src_file in enumerate(src_file_list):
if self.multilingual:
prepend_idx = idx
dataset = TranslationDataset(
dataset_src=str(Path(src_file).expanduser()),
dataset_tgt=str(Path(tgt_file_list[idx]).expanduser()),
tokens_in_batch=cfg.tokens_in_batch,
clean=cfg.get("clean", False),
max_seq_length=cfg.get("max_seq_length", 512),
min_seq_length=cfg.get("min_seq_length", 1),
max_seq_length_diff=cfg.get("max_seq_length_diff", 512),
max_seq_length_ratio=cfg.get("max_seq_length_ratio", 512),
cache_ids=cfg.get("cache_ids", False),
cache_data_per_node=cfg.get("cache_data_per_node", False),
use_cache=cfg.get("use_cache", False),
reverse_lang_direction=cfg.get("reverse_lang_direction",
False),
prepend_id=self.multilingual_ids[prepend_idx]
if self.multilingual else None,
)
dataset.batchify(self.encoder_tokenizer, self.decoder_tokenizer)
if cfg.shuffle:
sampler = pt_data.RandomSampler(dataset)
else:
sampler = pt_data.SequentialSampler(dataset)
dataloader = torch.utils.data.DataLoader(
dataset=dataset,
batch_size=1,
sampler=sampler,
num_workers=cfg.get("num_workers", 2),
pin_memory=cfg.get("pin_memory", False),
drop_last=cfg.get("drop_last", False),
)
dataloaders.append(dataloader)
return dataloaders
def setup_pre_and_post_processing_utils(self, source_lang, target_lang):
"""
Creates source and target processor objects for input and output pre/post-processing.
"""
self.source_processor, self.target_processor = None, None
if self.encoder_tokenizer_library == 'byte-level':
self.source_processor = ByteLevelProcessor()
elif (source_lang == 'en'
and target_lang == 'ja') or (source_lang == 'ja'
and target_lang == 'en'):
self.source_processor = EnJaProcessor(source_lang)
elif source_lang == 'zh':
self.source_processor = ChineseProcessor()
elif source_lang == 'hi':
self.source_processor = IndicProcessor(source_lang)
elif source_lang is not None and source_lang not in ['ja', 'zh', 'hi']:
self.source_processor = MosesProcessor(source_lang)
if self.decoder_tokenizer_library == 'byte-level':
self.target_processor = ByteLevelProcessor()
elif (source_lang == 'en'
and target_lang == 'ja') or (source_lang == 'ja'
and target_lang == 'en'):
self.target_processor = EnJaProcessor(target_lang)
elif target_lang == 'zh':
self.target_processor = ChineseProcessor()
elif target_lang == 'hi':
self.target_processor = IndicProcessor(target_lang)
elif target_lang is not None and target_lang not in ['ja', 'zh', 'hi']:
self.target_processor = MosesProcessor(target_lang)
return self.source_processor, self.target_processor
def ids_to_postprocessed_text(self,
beam_ids,
tokenizer,
processor,
filter_beam_ids=True):
if filter_beam_ids:
beam_ids = self.filter_predicted_ids(beam_ids)
translations = [
tokenizer.ids_to_text(tr) for tr in beam_ids.cpu().numpy()
]
if processor is not None:
translations = [
processor.detokenize(translation.split(' '))
for translation in translations
]
return translations
@torch.no_grad()
def batch_translate(self,
src: torch.LongTensor,
src_mask: torch.LongTensor,
return_beam_scores: bool = False):
"""
Translates a minibatch of inputs from source language to target language.
Args:
src: minibatch of inputs in the src language (batch x seq_len)
src_mask: mask tensor indicating elements to be ignored (batch x seq_len)
Returns:
translations: a list strings containing detokenized translations
inputs: a list of string containing detokenized inputs
"""
mode = self.training
try:
self.eval()
src_hiddens = self.encoder(input_ids=src, encoder_mask=src_mask)
best_translations = self.beam_search(
encoder_hidden_states=src_hiddens,
encoder_input_mask=src_mask,
return_beam_scores=return_beam_scores)
if return_beam_scores:
all_translations, scores, best_translations = best_translations
scores = scores.view(-1)
all_translations = self.ids_to_postprocessed_text(
all_translations,
self.decoder_tokenizer,
self.target_processor,
filter_beam_ids=True)
best_translations = self.ids_to_postprocessed_text(
best_translations,
self.decoder_tokenizer,
self.target_processor,
filter_beam_ids=True)
inputs = self.ids_to_postprocessed_text(src,
self.encoder_tokenizer,
self.source_processor,
filter_beam_ids=False)
finally:
self.train(mode=mode)
if return_beam_scores:
return inputs, all_translations, scores.data.cpu().numpy().tolist(
), best_translations
return inputs, best_translations
def prepare_inference_batch(self, text, prepend_ids=[], target=False):
inputs = []
processor = self.source_processor if not target else self.target_processor
tokenizer = self.encoder_tokenizer if not target else self.decoder_tokenizer
for txt in text:
if processor is not None:
txt = processor.normalize(txt)
txt = processor.tokenize(txt)
ids = tokenizer.text_to_ids(txt)
ids = prepend_ids + [tokenizer.bos_id] + ids + [tokenizer.eos_id]
inputs.append(ids)
max_len = max(len(txt) for txt in inputs)
src_ids_ = np.ones((len(inputs), max_len)) * tokenizer.pad_id
for i, txt in enumerate(inputs):
src_ids_[i][:len(txt)] = txt
src_mask = torch.FloatTensor(
(src_ids_ != tokenizer.pad_id)).to(self.device)
src = torch.LongTensor(src_ids_).to(self.device)
return src, src_mask
# TODO: We should drop source/target_lang arguments in favor of using self.src/tgt_language
@torch.no_grad()
def translate(self,
text: List[str],
source_lang: str = None,
target_lang: str = None,
return_beam_scores: bool = False) -> List[str]:
"""
Translates list of sentences from source language to target language.
Should be regular text, this method performs its own tokenization/de-tokenization
Args:
text: list of strings to translate
source_lang: if not None, corresponding MosesTokenizer and MosesPunctNormalizer will be run
target_lang: if not None, corresponding MosesDecokenizer will be run
Returns:
list of translated strings
"""
# __TODO__: This will reset both source and target processors even if you want to reset just one.
if source_lang is not None or target_lang is not None:
self.setup_pre_and_post_processing_utils(source_lang, target_lang)
mode = self.training
prepend_ids = []
if self.multilingual:
if source_lang is None or target_lang is None:
raise ValueError(
"Expect source_lang and target_lang to infer for multilingual model."
)
src_symbol = self.encoder_tokenizer.token_to_id('<' + source_lang +
'>')
tgt_symbol = self.encoder_tokenizer.token_to_id('<' + target_lang +
'>')
prepend_ids = [
src_symbol
if src_symbol in self.multilingual_ids else tgt_symbol
]
try:
self.eval()
src, src_mask = self.prepare_inference_batch(text, prepend_ids)
if return_beam_scores:
_, all_translations, scores, best_translations = self.batch_translate(
src, src_mask, return_beam_scores=True)
return all_translations, scores, best_translations
else:
_, translations = self.batch_translate(
src, src_mask, return_beam_scores=False)
finally:
self.train(mode=mode)
return translations
@classmethod
def list_available_models(cls) -> Optional[Dict[str, str]]:
"""
This method returns a list of pre-trained model which can be instantiated directly from NVIDIA's NGC cloud.
Returns:
List of available pre-trained models.
"""
result = []
model = PretrainedModelInfo(
pretrained_model_name="nmt_en_de_transformer12x2",
location=
"https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_en_de_transformer12x2/versions/1.0.0rc1/files/nmt_en_de_transformer12x2.nemo",
description=
"En->De translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_de_transformer12x2",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_de_en_transformer12x2",
location=
"https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_de_en_transformer12x2/versions/1.0.0rc1/files/nmt_de_en_transformer12x2.nemo",
description=
"De->En translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_de_en_transformer12x2",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_en_es_transformer12x2",
location=
"https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_en_es_transformer12x2/versions/1.0.0rc1/files/nmt_en_es_transformer12x2.nemo",
description=
"En->Es translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_es_transformer12x2",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_es_en_transformer12x2",
location=
"https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_es_en_transformer12x2/versions/1.0.0rc1/files/nmt_es_en_transformer12x2.nemo",
description=
"Es->En translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_es_en_transformer12x2",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_en_fr_transformer12x2",
location=
"https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_en_fr_transformer12x2/versions/1.0.0rc1/files/nmt_en_fr_transformer12x2.nemo",
description=
"En->Fr translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_fr_transformer12x2",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_fr_en_transformer12x2",
location=
"https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_fr_en_transformer12x2/versions/1.0.0rc1/files/nmt_fr_en_transformer12x2.nemo",
description=
"Fr->En translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_fr_en_transformer12x2",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_en_ru_transformer6x6",
location=
"https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_en_ru_transformer6x6/versions/1.0.0rc1/files/nmt_en_ru_transformer6x6.nemo",
description=
"En->Ru translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_ru_transformer6x6",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_ru_en_transformer6x6",
location=
"https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_ru_en_transformer6x6/versions/1.0.0rc1/files/nmt_ru_en_transformer6x6.nemo",
description=
"Ru->En translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_ru_en_transformer6x6",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_zh_en_transformer6x6",
location=
"https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_zh_en_transformer6x6/versions/1.0.0rc1/files/nmt_zh_en_transformer6x6.nemo",
description=
"Zh->En translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_zh_en_transformer6x6",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_en_zh_transformer6x6",
location=
"https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_en_zh_transformer6x6/versions/1.0.0rc1/files/nmt_en_zh_transformer6x6.nemo",
description=
"En->Zh translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_zh_transformer6x6",
)
result.append(model)
return result
class TransformerLMModel(ModelPT):
"""
Left-to-right Transformer language model.
"""
def __init__(self, cfg: DictConfig, trainer: Trainer = None):
# Get global rank and total number of GPU workers for IterableDataset partitioning, if applicable
self.world_size = 1
if trainer is not None:
self.world_size = trainer.num_nodes * trainer.num_gpus
cfg = model_utils.convert_model_config_to_dict_config(cfg)
cfg = model_utils.maybe_update_config_version(cfg)
# Instantiates tokenizer and register to be saved with NeMo Model archive
# After this call, ther will be self.tokenizer which can convert between tokens and token_ids.
self.setup_tokenizer(
tokenizer_name=cfg.tokenizer.get("tokenizer_name", "yttm"),
tokenizer_model=cfg.tokenizer.get("tokenizer_model", None),
vocab_file=cfg.tokenizer.get("vocab_file", None),
bpe_dropout=cfg.tokenizer.get("bpe_dropout", 0.0),
special_tokens=cfg.tokenizer.get("special_tokens", {}))
# init superclass
super().__init__(cfg=cfg, trainer=trainer)
# make vocabulary size divisible by 8 for fast fp16 training
vocab_size = 8 * math.ceil(self.tokenizer.vocab_size / 8)
# encoder from NeMo, Megatron-LM, or HuggingFace
encoder_cfg_dict = OmegaConf.to_container(cfg.get('encoder'))
encoder_cfg_dict['vocab_size'] = vocab_size
library = encoder_cfg_dict.pop('library', 'nemo')
model_name = encoder_cfg_dict.pop('model_name', None)
pretrained = encoder_cfg_dict.pop('pretrained', False)
self.encoder = get_transformer(
library=library,
model_name=model_name,
pretrained=pretrained,
config_dict=encoder_cfg_dict,
encoder=True,
pre_ln_final_layer_norm=encoder_cfg_dict.get(
'pre_ln_final_layer_norm',
encoder_cfg_dict.get('pre_ln', True)),
)
self.log_softmax = TokenClassifier(
hidden_size=self.encoder.hidden_size,
num_classes=vocab_size,
activation=cfg.head.activation,
log_softmax=cfg.head.log_softmax,
dropout=cfg.head.dropout,
use_transformer_init=cfg.head.use_transformer_init,
)
# tie weights of embedding and softmax matrices
self.log_softmax.mlp.layer0.weight = self.encoder.embedding.token_embedding.weight
std_init_range = 1 / self.encoder.hidden_size**0.5
# initialize weights if not using pretrained encoder
if not self._cfg.encoder.get('pretrained', False):
self.encoder.apply(lambda module: transformer_weights_init(
module, std_init_range))
self.log_softmax.apply(
lambda module: transformer_weights_init(module, std_init_range))
self.loss_fn = SmoothedCrossEntropyLoss(
pad_id=self.tokenizer.pad_id, label_smoothing=cfg.label_smoothing)
self.eval_loss_fn = SmoothedCrossEntropyLoss(
pad_id=self.tokenizer.pad_id)
self.eval_loss = GlobalAverageLossMetric(dist_sync_on_step=False,
take_avg_loss=True)
self.eval_ppl = SequencePerplexity()
@typecheck()
def forward(self, input_ids, attention_mask):
"""
No special modification required for Lightning, define it as you normally would
in the `nn.Module` in vanilla PyTorch.
"""
hidden_states = self.encoder(input_ids=input_ids,
encoder_mask=attention_mask)
log_probs = self.log_softmax(hidden_states=hidden_states)
return log_probs
def training_step(self, batch, batch_idx):
"""
Lightning calls this inside the training loop with the data from the training dataloader
passed in as `batch`.
"""
# forward pass
for i in range(len(batch)):
if batch[i].ndim == 3:
# Dataset returns already batched data and the first dimension of size 1
# added by DataLoader is excess.
batch[i] = batch[i].squeeze(dim=0)
ids, mask = batch
input_ids, labels = ids[:, :-1], ids[:, 1:]
input_mask = mask[:, :-1]
log_probs = self(input_ids=input_ids, attention_mask=input_mask)
train_loss = self.loss_fn(log_probs=log_probs, labels=labels)
tensorboard_logs = {
"train_loss": train_loss,
"lr": self._optimizer.param_groups[0]["lr"],
}
return {"loss": train_loss, "log": tensorboard_logs}
def eval_step(self, batch, batch_idx):
for i in range(len(batch)):
if batch[i].ndim == 3:
# Dataset returns already batched data and the first dimension of size 1
# added by DataLoader is excess.
batch[i] = batch[i].squeeze(dim=0)
ids, mask = batch
input_ids, labels = ids[:, :-1], ids[:, 1:]
input_mask, output_mask = mask[:, :-1], mask[:, 1:]
log_probs = self(input_ids=input_ids, attention_mask=input_mask)
eval_loss = self.eval_loss_fn(log_probs=log_probs, labels=labels)
self.eval_loss(loss=eval_loss,
num_measurements=log_probs.shape[0] *
log_probs.shape[1])
self.eval_ppl(log_probs=log_probs, labels=labels, mask=output_mask)
return {}
def test_step(self, batch, batch_idx):
return self.eval_step(batch, batch_idx)
def validation_step(self, batch, batch_idx):
"""
Lightning calls this inside the validation loop with the data from the validation dataloader
passed in as `batch`.
"""
return self.eval_step(batch, batch_idx)
def eval_epoch_end(self, outputs, mode):
eval_loss = self.eval_loss.compute()
eval_ppl = self.eval_ppl.compute()
self.log(f"{mode}_loss", eval_loss, sync_dist=True)
self.log(f"{mode}_PPL", eval_ppl, sync_dist=True)
dataset_name = "Validation" if mode == 'val' else "Test"
logging.info(
f"\n\n\n\n{dataset_name} PPL: {np.round(eval_ppl.item(), 2)}")
def validation_epoch_end(self, outputs):
"""
Called at the end of validation to aggregate outputs.
:param outputs: list of individual outputs of each validation step.
"""
self.eval_epoch_end(outputs, 'val')
self.eval_loss.reset()
self.eval_ppl.reset()
def test_epoch_end(self, outputs):
self.eval_epoch_end(outputs, 'test')
def setup_tokenizer(self,
tokenizer_name=None,
tokenizer_model=None,
vocab_file=None,
bpe_dropout=0.0,
special_tokens=None):
supported_tokenizers = ['yttm', 'huggingface', 'sentencepiece', 'word']
if tokenizer_name not in supported_tokenizers:
raise NotImplementedError(
f"Currently we only support tokenizers in {supported_tokenizers}."
)
self.tokenizer = get_tokenizer(
tokenizer_name=tokenizer_name,
tokenizer_model=self.register_artifact(
"cfg.tokenizer.tokenizer_model", tokenizer_model),
vocab_file=vocab_file,
bpe_dropout=bpe_dropout,
special_tokens=dict(special_tokens or {}),
use_fast=False,
)
def setup_training_data(self, train_data_config: Optional[DictConfig]):
self._train_dl = self._setup_dataloader_from_config(
cfg=train_data_config)
def setup_validation_data(self, val_data_config: Optional[DictConfig]):
self._validation_dl = self._setup_dataloader_from_config(
cfg=val_data_config)
def setup_test_data(self, test_data_config: Optional[DictConfig]):
self._test_dl = self._setup_dataloader_from_config(
cfg=test_data_config)
def _setup_dataloader_from_config(self, cfg: DictConfig, predict_last_k=0):
if cfg.get("use_tarred_dataset", False):
if cfg.get("metadata_file") is None:
raise FileNotFoundError(
"Trying to use tarred data set but could not find metadata path in config."
)
else:
metadata_file = cfg.get('metadata_file')
with open(metadata_file) as metadata_reader:
metadata = json.load(metadata_reader)
if cfg.get('tar_files') is None:
tar_files = metadata.get('tar_files')
if tar_files is not None:
logging.info(
f'Loading from tarred dataset {tar_files}')
else:
raise FileNotFoundError(
"Could not find tarred dataset in config or metadata."
)
else:
tar_files = cfg.get('tar_files')
if metadata.get('tar_files') is not None:
raise ValueError(
'Tar files specified in config and in metadata file. Tar files should only be specified once.'
)
dataset = TarredSentenceDataset(
text_tar_filepaths=tar_files,
metadata_path=metadata_file,
tokenizer=self.tokenizer,
shuffle_n=cfg.get("tar_shuffle_n", 100),
shard_strategy=cfg.get("shard_strategy", "scatter"),
global_rank=self.global_rank,
world_size=self.world_size,
)
return torch.utils.data.DataLoader(
dataset=dataset,
batch_size=1,
num_workers=cfg.get("num_workers", 2),
pin_memory=cfg.get("pin_memory", False),
drop_last=cfg.get("drop_last", False),
)
else:
dataset = SentenceDataset(
tokenizer=self.tokenizer,
dataset=cfg.file_name,
tokens_in_batch=cfg.tokens_in_batch,
clean=cfg.get("clean", False),
max_seq_length=cfg.get("max_seq_length", 512),
min_seq_length=cfg.get("min_seq_length", 1),
cache_ids=cfg.get("cache_ids", False),
)
if cfg.shuffle:
sampler = pt_data.RandomSampler(dataset)
else:
sampler = pt_data.SequentialSampler(dataset)
return torch.utils.data.DataLoader(
dataset=dataset,
batch_size=1,
sampler=sampler,
num_workers=cfg.get("num_workers", 2),
pin_memory=cfg.get("pin_memory", False),
drop_last=cfg.get("drop_last", False),
)
@classmethod
def list_available_models(cls) -> Optional[Dict[str, str]]:
pass
class MTEncDecModel(EncDecNLPModel):
"""
Encoder-decoder machine translation model.
"""
def __init__(self, cfg: MTEncDecModelConfig, trainer: Trainer = None):
cfg = model_utils.convert_model_config_to_dict_config(cfg)
# Get global rank and total number of GPU workers for IterableDataset partitioning, if applicable
# Global_rank and local_rank is set by LightningModule in Lightning 1.2.0
self.world_size = 1
if trainer is not None:
self.world_size = trainer.num_nodes * trainer.num_gpus
cfg = model_utils.maybe_update_config_version(cfg)
self.src_language: str = cfg.get("src_language", None)
self.tgt_language: str = cfg.get("tgt_language", None)
# Instantiates tokenizers and register to be saved with NeMo Model archive
# After this call, ther will be self.encoder_tokenizer and self.decoder_tokenizer
# Which can convert between tokens and token_ids for SRC and TGT languages correspondingly.
self.setup_enc_dec_tokenizers(
encoder_tokenizer_library=cfg.encoder_tokenizer.get('library', 'yttm'),
encoder_tokenizer_model=cfg.encoder_tokenizer.get('tokenizer_model'),
encoder_bpe_dropout=cfg.encoder_tokenizer.get('bpe_dropout', 0.0),
encoder_model_name=cfg.encoder.get('model_name') if hasattr(cfg.encoder, 'model_name') else None,
decoder_tokenizer_library=cfg.decoder_tokenizer.get('library', 'yttm'),
decoder_tokenizer_model=cfg.decoder_tokenizer.tokenizer_model,
decoder_bpe_dropout=cfg.decoder_tokenizer.get('bpe_dropout', 0.0),
decoder_model_name=cfg.decoder.get('model_name') if hasattr(cfg.decoder, 'model_name') else None,
)
# After this call, the model will have self.source_processor and self.target_processor objects
self.setup_pre_and_post_processing_utils(source_lang=self.src_language, target_lang=self.tgt_language)
# TODO: Why is this base constructor call so late in the game?
super().__init__(cfg=cfg, trainer=trainer)
# encoder from NeMo, Megatron-LM, or HuggingFace
encoder_cfg_dict = OmegaConf.to_container(cfg.get('encoder'))
encoder_cfg_dict['vocab_size'] = self.encoder_vocab_size
library = encoder_cfg_dict.pop('library', 'nemo')
model_name = encoder_cfg_dict.pop('model_name', None)
pretrained = encoder_cfg_dict.pop('pretrained', False)
self.encoder = get_transformer(
library=library, model_name=model_name, pretrained=pretrained, config_dict=encoder_cfg_dict, encoder=True,
)
# decoder from NeMo, Megatron-LM, or HuggingFace
decoder_cfg_dict = OmegaConf.to_container(cfg.get('decoder'))
decoder_cfg_dict['vocab_size'] = self.decoder_vocab_size
library = decoder_cfg_dict.pop('library', 'nemo')
model_name = decoder_cfg_dict.pop('model_name', None)
pretrained = decoder_cfg_dict.pop('pretrained', False)
decoder_cfg_dict['hidden_size'] = self.encoder.hidden_size
self.decoder = get_transformer(
library=library, model_name=model_name, pretrained=pretrained, config_dict=decoder_cfg_dict, encoder=False,
)
self.log_softmax = TokenClassifier(
hidden_size=self.decoder.hidden_size,
num_classes=self.decoder_vocab_size,
activation=cfg.head.activation,
log_softmax=cfg.head.log_softmax,
dropout=cfg.head.dropout,
use_transformer_init=cfg.head.use_transformer_init,
)
self.beam_search = BeamSearchSequenceGenerator(
embedding=self.decoder.embedding,
decoder=self.decoder.decoder,
log_softmax=self.log_softmax,
max_sequence_length=self.decoder.max_sequence_length,
beam_size=cfg.beam_size,
bos=self.decoder_tokenizer.bos_id,
pad=self.decoder_tokenizer.pad_id,
eos=self.decoder_tokenizer.eos_id,
len_pen=cfg.len_pen,
max_delta_length=cfg.max_generation_delta,
)
# tie weights of embedding and softmax matrices
self.log_softmax.mlp.layer0.weight = self.decoder.embedding.token_embedding.weight
# TODO: encoder and decoder with different hidden size?
std_init_range = 1 / self.encoder.hidden_size ** 0.5
# initialize weights if not using pretrained encoder/decoder
if not self._cfg.encoder.get('pretrained', False):
self.encoder.apply(lambda module: transformer_weights_init(module, std_init_range))
if not self._cfg.decoder.get('pretrained', False):
self.decoder.apply(lambda module: transformer_weights_init(module, std_init_range))
self.log_softmax.apply(lambda module: transformer_weights_init(module, std_init_range))
self.loss_fn = SmoothedCrossEntropyLoss(
pad_id=self.decoder_tokenizer.pad_id, label_smoothing=cfg.label_smoothing
)
self.eval_loss = GlobalAverageLossMetric(dist_sync_on_step=False, take_avg_loss=True)
def filter_predicted_ids(self, ids):
ids[ids >= self.decoder_tokenizer.vocab_size] = self.decoder_tokenizer.unk_id
return ids
@typecheck()
def forward(self, src, src_mask, tgt, tgt_mask):
src_hiddens = self.encoder(input_ids=src, encoder_mask=src_mask)
tgt_hiddens = self.decoder(
input_ids=tgt, decoder_mask=tgt_mask, encoder_embeddings=src_hiddens, encoder_mask=src_mask
)
log_probs = self.log_softmax(hidden_states=tgt_hiddens)
return log_probs
def training_step(self, batch, batch_idx):
"""
Lightning calls this inside the training loop with the data from the training dataloader
passed in as `batch`.
"""
# forward pass
for i in range(len(batch)):
if batch[i].ndim == 3:
# Dataset returns already batched data and the first dimension of size 1 added by DataLoader
# is excess.
batch[i] = batch[i].squeeze(dim=0)
src_ids, src_mask, tgt_ids, tgt_mask, labels = batch
log_probs = self(src_ids, src_mask, tgt_ids, tgt_mask)
train_loss = self.loss_fn(log_probs=log_probs, labels=labels)
tensorboard_logs = {
'train_loss': train_loss,
'lr': self._optimizer.param_groups[0]['lr'],
}
return {'loss': train_loss, 'log': tensorboard_logs}
def eval_step(self, batch, batch_idx, mode):
for i in range(len(batch)):
if batch[i].ndim == 3:
# Dataset returns already batched data and the first dimension of size 1 added by DataLoader
# is excess.
batch[i] = batch[i].squeeze(dim=0)
src_ids, src_mask, tgt_ids, tgt_mask, labels = batch
log_probs = self(src_ids, src_mask, tgt_ids, tgt_mask)
# this will run encoder twice -- TODO: potentially fix
_, translations = self.batch_translate(src=src_ids, src_mask=src_mask)
eval_loss = self.loss_fn(log_probs=log_probs, labels=labels)
self.eval_loss(loss=eval_loss, num_measurements=log_probs.shape[0] * log_probs.shape[1])
np_tgt = tgt_ids.detach().cpu().numpy()
ground_truths = [self.decoder_tokenizer.ids_to_text(tgt) for tgt in np_tgt]
ground_truths = [self.target_processor.detokenize(tgt.split(' ')) for tgt in ground_truths]
num_non_pad_tokens = np.not_equal(np_tgt, self.decoder_tokenizer.pad_id).sum().item()
return {
'translations': translations,
'ground_truths': ground_truths,
'num_non_pad_tokens': num_non_pad_tokens,
}
def test_step(self, batch, batch_idx):
return self.eval_step(batch, batch_idx, 'test')
@rank_zero_only
def log_param_stats(self):
for name, p in self.named_parameters():
if p.requires_grad:
self.trainer.logger.experiment.add_histogram(name + '_hist', p, global_step=self.global_step)
self.trainer.logger.experiment.add_scalars(
name,
{'mean': p.mean(), 'stddev': p.std(), 'max': p.max(), 'min': p.min()},
global_step=self.global_step,
)
def validation_step(self, batch, batch_idx):
"""
Lightning calls this inside the validation loop with the data from the validation dataloader
passed in as `batch`.
"""
return self.eval_step(batch, batch_idx, 'val')
def eval_epoch_end(self, outputs, mode):
eval_loss = self.eval_loss.compute()
translations = list(itertools.chain(*[x['translations'] for x in outputs]))
ground_truths = list(itertools.chain(*[x['ground_truths'] for x in outputs]))
assert len(translations) == len(ground_truths)
if self.tgt_language in ['ja']:
sacre_bleu = corpus_bleu(translations, [ground_truths], tokenize="ja-mecab")
elif self.tgt_language in ['zh']:
sacre_bleu = corpus_bleu(translations, [ground_truths], tokenize="zh")
else:
sacre_bleu = corpus_bleu(translations, [ground_truths], tokenize="13a")
dataset_name = "Validation" if mode == 'val' else "Test"
logging.info(f"\n\n\n\n{dataset_name} set size: {len(translations)}")
logging.info(f"{dataset_name} Sacre BLEU = {sacre_bleu.score}")
logging.info(f"{dataset_name} TRANSLATION EXAMPLES:".upper())
for i in range(0, 3):
ind = random.randint(0, len(translations) - 1)
logging.info(" " + '\u0332'.join(f"EXAMPLE {i}:"))
logging.info(f" Prediction: {translations[ind]}")
logging.info(f" Ground Truth: {ground_truths[ind]}")
ans = {f"{mode}_loss": eval_loss, f"{mode}_sacreBLEU": sacre_bleu.score}
ans['log'] = dict(ans)
return ans
def validation_epoch_end(self, outputs):
"""
Called at the end of validation to aggregate outputs.
:param outputs: list of individual outputs of each validation step.
"""
self.log_dict(self.eval_epoch_end(outputs, 'val'), sync_dist=True)
def test_epoch_end(self, outputs):
return self.eval_epoch_end(outputs, 'test')
def setup_enc_dec_tokenizers(
self,
encoder_tokenizer_library=None,
encoder_tokenizer_model=None,
encoder_bpe_dropout=0.0,
encoder_model_name=None,
decoder_tokenizer_library=None,
decoder_tokenizer_model=None,
decoder_bpe_dropout=0.0,
decoder_model_name=None,
):
supported_tokenizers = ['yttm', 'huggingface', 'sentencepiece']
if (
encoder_tokenizer_library not in supported_tokenizers
or decoder_tokenizer_library not in supported_tokenizers
):
raise NotImplementedError(f"Currently we only support tokenizers in {supported_tokenizers}.")
self.encoder_tokenizer = get_nmt_tokenizer(
library=encoder_tokenizer_library,
tokenizer_model=self.register_artifact("cfg.encoder_tokenizer.tokenizer_model", encoder_tokenizer_model),
bpe_dropout=encoder_bpe_dropout,
model_name=encoder_model_name,
vocab_file=None,
special_tokens=None,
use_fast=False,
)
self.decoder_tokenizer = get_nmt_tokenizer(
library=decoder_tokenizer_library,
tokenizer_model=self.register_artifact("cfg.decoder_tokenizer.tokenizer_model", decoder_tokenizer_model),
bpe_dropout=decoder_bpe_dropout,
model_name=decoder_model_name,
vocab_file=None,
special_tokens=None,
use_fast=False,
)
def setup_training_data(self, train_data_config: Optional[DictConfig]):
self._train_dl = self._setup_dataloader_from_config(cfg=train_data_config)
def setup_validation_data(self, val_data_config: Optional[DictConfig]):
self._validation_dl = self._setup_dataloader_from_config(cfg=val_data_config)
def setup_test_data(self, test_data_config: Optional[DictConfig]):
self._test_dl = self._setup_dataloader_from_config(cfg=test_data_config)
def _setup_dataloader_from_config(self, cfg: DictConfig):
if cfg.get("load_from_cached_dataset", False):
logging.info('Loading from cached dataset %s' % (cfg.src_file_name))
if cfg.src_file_name != cfg.tgt_file_name:
raise ValueError("src must be equal to target for cached dataset")
dataset = pickle.load(open(cfg.src_file_name, 'rb'))
dataset.reverse_lang_direction = cfg.get("reverse_lang_direction", False)
elif cfg.get("use_tarred_dataset", False):
if cfg.get("metadata_file") is None:
raise FileNotFoundError("Trying to use tarred data set but could not find metadata path in config.")
else:
metadata_file = cfg.get('metadata_file')
with open(metadata_file) as metadata_reader:
metadata = json.load(metadata_reader)
if cfg.get('tar_files') is None:
tar_files = metadata.get('tar_files')
if tar_files is not None:
logging.info(f'Loading from tarred dataset {tar_files}')
else:
raise FileNotFoundError("Could not find tarred dataset in config or metadata.")
else:
tar_files = cfg.get('tar_files')
if metadata_file.get('tar_files') is not None:
raise ValueError(
'Tar files specified in config and in metadata file. Tar files should only be specified once.'
)
dataset = TarredTranslationDataset(
text_tar_filepaths=tar_files,
metadata_path=metadata_file,
encoder_tokenizer=self.encoder_tokenizer,
decoder_tokenizer=self.decoder_tokenizer,
shuffle_n=cfg.get("tar_shuffle_n", 100),
shard_strategy=cfg.get("shard_strategy", "scatter"),
global_rank=self.global_rank,
world_size=self.world_size,
reverse_lang_direction=cfg.get("reverse_lang_direction", False),
)
return torch.utils.data.DataLoader(
dataset=dataset,
batch_size=1,
num_workers=cfg.get("num_workers", 2),
pin_memory=cfg.get("pin_memory", False),
drop_last=cfg.get("drop_last", False),
)
else:
dataset = TranslationDataset(
dataset_src=str(Path(cfg.src_file_name).expanduser()),
dataset_tgt=str(Path(cfg.tgt_file_name).expanduser()),
tokens_in_batch=cfg.tokens_in_batch,
clean=cfg.get("clean", False),
max_seq_length=cfg.get("max_seq_length", 512),
min_seq_length=cfg.get("min_seq_length", 1),
max_seq_length_diff=cfg.get("max_seq_length_diff", 512),
max_seq_length_ratio=cfg.get("max_seq_length_ratio", 512),
cache_ids=cfg.get("cache_ids", False),
cache_data_per_node=cfg.get("cache_data_per_node", False),
use_cache=cfg.get("use_cache", False),
reverse_lang_direction=cfg.get("reverse_lang_direction", False),
)
dataset.batchify(self.encoder_tokenizer, self.decoder_tokenizer)
if cfg.shuffle:
sampler = pt_data.RandomSampler(dataset)
else:
sampler = pt_data.SequentialSampler(dataset)
return torch.utils.data.DataLoader(
dataset=dataset,
batch_size=1,
sampler=sampler,
num_workers=cfg.get("num_workers", 2),
pin_memory=cfg.get("pin_memory", False),
drop_last=cfg.get("drop_last", False),
)
def setup_pre_and_post_processing_utils(self, source_lang, target_lang):
"""
Creates source and target processor objects for input and output pre/post-processing.
"""
self.source_processor, self.target_processor = None, None
if (source_lang == 'en' and target_lang == 'ja') or (source_lang == 'ja' and target_lang == 'en'):
self.source_processor = EnJaProcessor(source_lang)
self.target_processor = EnJaProcessor(target_lang)
else:
if source_lang == 'zh':
self.source_processor = ChineseProcessor()
if target_lang == 'zh':
self.target_processor = ChineseProcessor()
if source_lang is not None and source_lang not in ['ja', 'zh']:
self.source_processor = MosesProcessor(source_lang)
if target_lang is not None and target_lang not in ['ja', 'zh']:
self.target_processor = MosesProcessor(target_lang)
@torch.no_grad()
def batch_translate(
self, src: torch.LongTensor, src_mask: torch.LongTensor,
):
"""
Translates a minibatch of inputs from source language to target language.
Args:
src: minibatch of inputs in the src language (batch x seq_len)
src_mask: mask tensor indicating elements to be ignored (batch x seq_len)
Returns:
translations: a list strings containing detokenized translations
inputs: a list of string containing detokenized inputs
"""
mode = self.training
try:
self.eval()
src_hiddens = self.encoder(input_ids=src, encoder_mask=src_mask)
beam_results = self.beam_search(encoder_hidden_states=src_hiddens, encoder_input_mask=src_mask)
beam_results = self.filter_predicted_ids(beam_results)
translations = [self.decoder_tokenizer.ids_to_text(tr) for tr in beam_results.cpu().numpy()]
inputs = [self.encoder_tokenizer.ids_to_text(inp) for inp in src.cpu().numpy()]
if self.target_processor is not None:
translations = [
self.target_processor.detokenize(translation.split(' ')) for translation in translations
]
if self.source_processor is not None:
inputs = [self.source_processor.detokenize(item.split(' ')) for item in inputs]
finally:
self.train(mode=mode)
return inputs, translations
# TODO: We should drop source/target_lang arguments in favor of using self.src/tgt_language
@torch.no_grad()
def translate(self, text: List[str], source_lang: str = None, target_lang: str = None) -> List[str]:
"""
Translates list of sentences from source language to target language.
Should be regular text, this method performs its own tokenization/de-tokenization
Args:
text: list of strings to translate
source_lang: if not None, corresponding MosesTokenizer and MosesPunctNormalizer will be run
target_lang: if not None, corresponding MosesDecokenizer will be run
Returns:
list of translated strings
"""
# __TODO__: This will reset both source and target processors even if you want to reset just one.
if source_lang is not None or target_lang is not None:
self.setup_pre_and_post_processing_utils(source_lang, target_lang)
mode = self.training
try:
self.eval()
inputs = []
for txt in text:
if self.source_processor is not None:
txt = self.source_processor.normalize(txt)
txt = self.source_processor.tokenize(txt)
ids = self.encoder_tokenizer.text_to_ids(txt)
ids = [self.encoder_tokenizer.bos_id] + ids + [self.encoder_tokenizer.eos_id]
inputs.append(ids)
max_len = max(len(txt) for txt in inputs)
src_ids_ = np.ones((len(inputs), max_len)) * self.encoder_tokenizer.pad_id
for i, txt in enumerate(inputs):
src_ids_[i][: len(txt)] = txt
src_mask = torch.FloatTensor((src_ids_ != self.encoder_tokenizer.pad_id)).to(self.device)
src = torch.LongTensor(src_ids_).to(self.device)
_, translations = self.batch_translate(src, src_mask)
finally:
self.train(mode=mode)
return translations
@classmethod
def list_available_models(cls) -> Optional[Dict[str, str]]:
"""
This method returns a list of pre-trained model which can be instantiated directly from NVIDIA's NGC cloud.
Returns:
List of available pre-trained models.
"""
result = []
model = PretrainedModelInfo(
pretrained_model_name="nmt_en_de_transformer12x2",
location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_en_de_transformer12x2/versions/1.0.0rc1/files/nmt_en_de_transformer12x2.nemo",
description="En->De translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_de_transformer12x2",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_de_en_transformer12x2",
location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_de_en_transformer12x2/versions/1.0.0rc1/files/nmt_de_en_transformer12x2.nemo",
description="De->En translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_de_en_transformer12x2",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_en_es_transformer12x2",
location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_en_es_transformer12x2/versions/1.0.0rc1/files/nmt_en_es_transformer12x2.nemo",
description="En->Es translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_es_transformer12x2",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_es_en_transformer12x2",
location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_es_en_transformer12x2/versions/1.0.0rc1/files/nmt_es_en_transformer12x2.nemo",
description="Es->En translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_es_en_transformer12x2",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_en_fr_transformer12x2",
location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_en_fr_transformer12x2/versions/1.0.0rc1/files/nmt_en_fr_transformer12x2.nemo",
description="En->Fr translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_fr_transformer12x2",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_fr_en_transformer12x2",
location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_fr_en_transformer12x2/versions/1.0.0rc1/files/nmt_fr_en_transformer12x2.nemo",
description="Fr->En translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_fr_en_transformer12x2",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_en_ru_transformer6x6",
location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_en_ru_transformer6x6/versions/1.0.0rc1/files/nmt_en_ru_transformer6x6.nemo",
description="En->Ru translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_ru_transformer6x6",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_ru_en_transformer6x6",
location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_ru_en_transformer6x6/versions/1.0.0rc1/files/nmt_ru_en_transformer6x6.nemo",
description="Ru->En translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_ru_en_transformer6x6",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_zh_en_transformer6x6",
location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_zh_en_transformer6x6/versions/1.0.0rc1/files/nmt_zh_en_transformer6x6.nemo",
description="Zh->En translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_zh_en_transformer6x6",
)
result.append(model)
model = PretrainedModelInfo(
pretrained_model_name="nmt_en_zh_transformer6x6",
location="https://api.ngc.nvidia.com/v2/models/nvidia/nemo/nmt_en_zh_transformer6x6/versions/1.0.0rc1/files/nmt_en_zh_transformer6x6.nemo",
description="En->Zh translation model. See details here: https://ngc.nvidia.com/catalog/models/nvidia:nemo:nmt_en_zh_transformer6x6",
)
result.append(model)
return result