Smaller BERT Sentence Embeddings (L-12_H-128_A-2)

Description

This is one of the smaller BERT models referenced in Well-Read Students Learn Better: On the Importance of Pre-training Compact Models. The smaller BERT models are intended for environments with restricted computational resources. They can be fine-tuned in the same manner as the original BERT models. However, they are most effective in the context of knowledge distillation, where the fine-tuning labels are produced by a larger and more accurate teacher.

Live Demo Open in Colab Download Copy S3 URI

How to use

...
embeddings = BertSentenceEmbeddings.pretrained("sent_small_bert_L12_128", "en") \
.setInputCols("sentence") \
.setOutputCol("sentence_embeddings")
nlp_pipeline = Pipeline(stages=[document_assembler, sentence_detector, embeddings])
pipeline_model = nlp_pipeline.fit(spark.createDataFrame([[""]]).toDF("text"))
result = pipeline_model.transform(spark.createDataFrame([['I hate cancer', "Antibiotics aren't painkiller"]], ["text"]))

...
val embeddings = BertSentenceEmbeddings.pretrained("sent_small_bert_L12_128", "en")
.setInputCols("sentence")
.setOutputCol("sentence_embeddings")
val pipeline = new Pipeline().setStages(Array(document_assembler, sentence_detector, embeddings))
val data = Seq("I hate cancer, "Antibiotics aren't painkiller").toDF("text")
val result = pipeline.fit(data).transform(data)

import nlu

text = ["I hate cancer", "Antibiotics aren't painkiller"]
embeddings_df = nlu.load('en.embed_sentence.small_bert_L12_128').predict(text, output_level='sentence')
embeddings_df

Results

	en_embed_sentence_small_bert_L12_128_embeddings	      sentence
		
	[-0.3747739791870117, -0.28460437059402466, 0.... 	I hate cancer
	[0.9055836200714111, -0.41459062695503235, 0.0... 	Antibiotics aren't painkiller

Model Information

Data Source

The model is imported from https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-12_H-128_A-2/1