This repo contains a TensorFlow 2.0 Keras implementation of google-research/bert with support for loading of the original pre-trained weights, and producing activations numerically identical to the one calculated by the original model.
ALBERT and adapter-BERT are also supported by setting the corresponding
configuration parameters (shared_layer=True
, embedding_size
for ALBERT
and adapter_size
for adapter-BERT). Setting both will result in an adapter-ALBERT
by sharing the BERT parameters across all layers while adapting every layer with layer specific adapter.
The implementation is build from scratch using only basic tensorflow operations, following the code in google-research/bert/modeling.py (but skipping dead code and applying some simplifications). It also utilizes kpe/params-flow to reduce common Keras boilerplate code (related to passing model and layer configuration arguments).
bert-for-tf2 should work with both TensorFlow 2.0 and TensorFlow 1.14 or newer.
- 06.Apr.2020 - using latest
py-params
introducingWithParams
base forLayer
andModel
. See news in kpe/py-params for how to update (_construct()
signature has change and requires callingsuper().__construct()
).- 06.Jan.2020 - support for loading the tar format weights from google-research/ALBERT.
- 18.Nov.2019 - ALBERT tokenization added (make sure to import as
from bert import albert_tokenization
orfrom bert import bert_tokenization
).- 08.Nov.2019 - using v2 per default when loading the TFHub/albert weights of google-research/ALBERT.
- 05.Nov.2019 - minor ALBERT word embeddings refactoring (
word_embeddings_2
->word_embeddings_projector
) and related parameter freezing fixes.- 04.Nov.2019 - support for extra (task specific) token embeddings using negative token ids.
- 29.Oct.2019 - support for loading of the pre-trained ALBERT weights released by google-research/ALBERT at TFHub/albert.
- 11.Oct.2019 - support for loading of the pre-trained ALBERT weights released by brightmart/albert_zh ALBERT for Chinese.
- 10.Oct.2019 - support for ALBERT through the
shared_layer=True
andembedding_size=128
params.- 03.Sep.2019 - walkthrough on fine tuning with adapter-BERT and storing the fine tuned fraction of the weights in a separate checkpoint (see
tests/test_adapter_finetune.py
).- 02.Sep.2019 - support for extending the token type embeddings of a pre-trained model by returning the mismatched weights in
load_stock_weights()
(seetests/test_extend_segments.py
).- 25.Jul.2019 - there are now two colab notebooks under
examples/
showing how to fine-tune an IMDB Movie Reviews sentiment classifier from pre-trained BERT weights using an adapter-BERT model architecture on a GPU or TPU in Google Colab.- 28.Jun.2019 - v.0.3.0 supports adapter-BERT (google-research/adapter-bert) for "Parameter-Efficient Transfer Learning for NLP", i.e. fine-tuning small overlay adapter layers over BERT's transformer encoders without changing the frozen BERT weights.
MIT. See License File.
bert-for-tf2
is on the Python Package Index (PyPI):
pip install bert-for-tf2
BERT in bert-for-tf2 is implemented as a Keras layer. You could instantiate it like this:
from bert import BertModelLayer
l_bert = BertModelLayer(**BertModelLayer.Params(
vocab_size = 16000, # embedding params
use_token_type = True,
use_position_embeddings = True,
token_type_vocab_size = 2,
num_layers = 12, # transformer encoder params
hidden_size = 768,
hidden_dropout = 0.1,
intermediate_size = 4*768,
intermediate_activation = "gelu",
adapter_size = None, # see arXiv:1902.00751 (adapter-BERT)
shared_layer = False, # True for ALBERT (arXiv:1909.11942)
embedding_size = None, # None for BERT, wordpiece embedding size for ALBERT
name = "bert" # any other Keras layer params
))
or by using the bert_config.json
from a pre-trained google model:
import bert
model_dir = ".models/uncased_L-12_H-768_A-12"
bert_params = bert.params_from_pretrained_ckpt(model_dir)
l_bert = bert.BertModelLayer.from_params(bert_params, name="bert")
now you can use the BERT layer in your Keras model like this:
from tensorflow import keras
max_seq_len = 128
l_input_ids = keras.layers.Input(shape=(max_seq_len,), dtype='int32')
l_token_type_ids = keras.layers.Input(shape=(max_seq_len,), dtype='int32')
# using the default token_type/segment id 0
output = l_bert(l_input_ids) # output: [batch_size, max_seq_len, hidden_size]
model = keras.Model(inputs=l_input_ids, outputs=output)
model.build(input_shape=(None, max_seq_len))
# provide a custom token_type/segment id as a layer input
output = l_bert([l_input_ids, l_token_type_ids]) # [batch_size, max_seq_len, hidden_size]
model = keras.Model(inputs=[l_input_ids, l_token_type_ids], outputs=output)
model.build(input_shape=[(None, max_seq_len), (None, max_seq_len)])
if you choose to use adapter-BERT by setting the adapter_size parameter, you would also like to freeze all the original BERT layers by calling:
l_bert.apply_adapter_freeze()
and once the model has been build or compiled, the original pre-trained weights can be loaded in the BERT layer:
import bert
bert_ckpt_file = os.path.join(model_dir, "bert_model.ckpt")
bert.load_stock_weights(l_bert, bert_ckpt_file)
N.B. see tests/test_bert_activations.py for a complete example.
- How to use BERT with the google-research/bert pre-trained weights?
model_name = "uncased_L-12_H-768_A-12"
model_dir = bert.fetch_google_bert_model(model_name, ".models")
model_ckpt = os.path.join(model_dir, "bert_model.ckpt")
bert_params = bert.params_from_pretrained_ckpt(model_dir)
l_bert = bert.BertModelLayer.from_params(bert_params, name="bert")
# use in Keras Model here, and call model.build()
bert.load_bert_weights(l_bert, model_ckpt) # should be called after model.build()
- How to use ALBERT with the google-research/ALBERT pre-trained weights (fetching from TFHub)?
see tests/nonci/test_load_pretrained_weights.py:
model_name = "albert_base"
model_dir = bert.fetch_tfhub_albert_model(model_name, ".models")
model_params = bert.albert_params(model_name)
l_bert = bert.BertModelLayer.from_params(model_params, name="albert")
# use in Keras Model here, and call model.build()
bert.load_albert_weights(l_bert, albert_dir) # should be called after model.build()
- How to use ALBERT with the google-research/ALBERT pre-trained weights (non TFHub)?
see tests/nonci/test_load_pretrained_weights.py:
model_name = "albert_base_v2"
model_dir = bert.fetch_google_albert_model(model_name, ".models")
model_ckpt = os.path.join(albert_dir, "model.ckpt-best")
model_params = bert.albert_params(model_dir)
l_bert = bert.BertModelLayer.from_params(model_params, name="albert")
# use in Keras Model here, and call model.build()
bert.load_albert_weights(l_bert, model_ckpt) # should be called after model.build()
- How to use ALBERT with the brightmart/albert_zh pre-trained weights?
see tests/nonci/test_albert.py:
model_name = "albert_base"
model_dir = bert.fetch_brightmart_albert_model(model_name, ".models")
model_ckpt = os.path.join(model_dir, "albert_model.ckpt")
bert_params = bert.params_from_pretrained_ckpt(model_dir)
l_bert = bert.BertModelLayer.from_params(bert_params, name="bert")
# use in a Keras Model here, and call model.build()
bert.load_albert_weights(l_bert, model_ckpt) # should be called after model.build()
- How to tokenize the input for the google-research/bert models?
do_lower_case = not (model_name.find("cased") == 0 or model_name.find("multi_cased") == 0)
bert.bert_tokenization.validate_case_matches_checkpoint(do_lower_case, model_ckpt)
vocab_file = os.path.join(model_dir, "vocab.txt")
tokenizer = bert.bert_tokenization.FullTokenizer(vocab_file, do_lower_case)
tokens = tokenizer.tokenize("Hello, BERT-World!")
token_ids = tokenizer.convert_tokens_to_ids(tokens)
- How to tokenize the input for brightmart/albert_zh?
import params_flow pf
# fetch the vocab file
albert_zh_vocab_url = "https://raw.githubusercontent.com/brightmart/albert_zh/master/albert_config/vocab.txt"
vocab_file = pf.utils.fetch_url(albert_zh_vocab_url, model_dir)
tokenizer = bert.albert_tokenization.FullTokenizer(vocab_file)
tokens = tokenizer.tokenize("你好世界")
token_ids = tokenizer.convert_tokens_to_ids(tokens)
- How to tokenize the input for the google-research/ALBERT models?
import sentencepiece as spm
spm_model = os.path.join(model_dir, "assets", "30k-clean.model")
sp = spm.SentencePieceProcessor()
sp.load(spm_model)
do_lower_case = True
processed_text = bert.albert_tokenization.preprocess_text("Hello, World!", lower=do_lower_case)
token_ids = bert.albert_tokenization.encode_ids(sp, processed_text)
- How to tokenize the input for the Chinese google-research/ALBERT models?
import bert
vocab_file = os.path.join(model_dir, "vocab.txt")
tokenizer = bert.albert_tokenization.FullTokenizer(vocab_file=vocab_file)
tokens = tokenizer.tokenize(u"你好世界")
token_ids = tokenizer.convert_tokens_to_ids(tokens)
- BERT - BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
- adapter-BERT - adapter-BERT: Parameter-Efficient Transfer Learning for NLP
- ALBERT - ALBERT: A Lite BERT for Self-Supervised Learning of Language Representations
- google-research/bert - the original BERT implementation
- google-research/ALBERT - the original ALBERT implementation by Google
- google-research/albert(old) - the old location of the original ALBERT implementation by Google
- brightmart/albert_zh - pre-trained ALBERT weights for Chinese
- kpe/params-flow - A Keras coding style for reducing Keras boilerplate code in custom layers by utilizing kpe/py-params