Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[BUGFIX] Megatron in NMT was setting vocab_file to None #2417

Merged
merged 3 commits into from
Jun 29, 2021
Merged

[BUGFIX] Megatron in NMT was setting vocab_file to None #2417

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
18849cb
make vocab_file configurable for megatron in nmt
ericharper Jun 28, 2021
f9c1e2b
update docs
ericharper Jun 28, 2021
c3c566c
update docs
ericharper Jun 28, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 11 additions & 0 deletions docs/source/nlp/machine_translation.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -561,6 +561,17 @@ To train a Megatron 345M BERT, we would use
model.encoder.num_layers=24 \
model.encoder.max_position_embeddings=512 \

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

.. code::

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
----------
Expand Down
4 changes: 3 additions & 1 deletion nemo/collections/nlp/models/machine_translation/mt_enc_dec_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,6 +80,7 @@ def __init__(self, cfg: MTEncDecModelConfig, trainer: Trainer = None):
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_tokenizer_vocab_file=cfg.encoder_tokenizer.get('vocab_file', 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)
Expand Down Expand Up @@ -379,6 +380,7 @@ def setup_enc_dec_tokenizers(
encoder_tokenizer_model=None,
encoder_bpe_dropout=0.0,
encoder_model_name=None,
encoder_tokenizer_vocab_file=None,
decoder_tokenizer_library=None,
decoder_tokenizer_model=None,
decoder_bpe_dropout=0.0,
Expand All @@ -397,7 +399,7 @@ def setup_enc_dec_tokenizers(
tokenizer_model=self.register_artifact("encoder_tokenizer.tokenizer_model", encoder_tokenizer_model),
bpe_dropout=encoder_bpe_dropout,
model_name=encoder_model_name,
vocab_file=None,
vocab_file=self.register_artifact("encoder_tokenizer.vocab_file", encoder_tokenizer_vocab_file),
special_tokens=None,
use_fast=False,
)
Expand Down
5 changes: 4 additions & 1 deletion nemo/collections/nlp/modules/common/tokenizer_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -95,6 +95,9 @@ def get_tokenizer(
elif tokenizer_name == 'char':
return CharTokenizer(vocab_file=vocab_file, **special_tokens_dict)

logging.info(
f"Getting HuggingFace AutoTokenizer with pretrained_model_name: {tokenizer_name}, vocab_file: {vocab_file}, special_tokens_dict: {special_tokens_dict}, and use_fast: {use_fast}"
)
return AutoTokenizer(
pretrained_model_name=tokenizer_name, vocab_file=vocab_file, **special_tokens_dict, use_fast=use_fast
)
Expand Down Expand Up @@ -140,7 +143,7 @@ def get_nmt_tokenizer(
)
elif library == 'megatron':
logging.info(
f'Getting Megatron tokenizer with pretrained model name: {model_name} and custom vocab file: {vocab_file}'
f'Getting Megatron tokenizer for pretrained model name: {model_name} and custom vocab file: {vocab_file}'
)
return get_tokenizer(tokenizer_name=model_name, vocab_file=vocab_file)
else:
Expand Down