Smaller BERT 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.

Download

How to use

...
embeddings = BertEmbeddings.pretrained("small_bert_L12_128", "en") \
      .setInputCols("sentence", "token") \
      .setOutputCol("embeddings")
nlp_pipeline = Pipeline(stages=[document_assembler, sentence_detector, tokenizer, embeddings])
pipeline_model = nlp_pipeline.fit(spark.createDataFrame([[""]]).toDF("text"))
result = pipeline_model.transform(spark.createDataFrame(pd.DataFrame({"text": ["I love NLP"]})))
...
val embeddings = BertEmbeddings.pretrained("small_bert_L12_128", "en")
      .setInputCols("sentence", "token")
      .setOutputCol("embeddings")
val pipeline = new Pipeline().setStages(Array(document_assembler, sentence_detector, tokenizer, embeddings))
val result = pipeline.fit(Seq.empty["I love NLP"].toDS.toDF("text")).transform(data)
import nlu

text = ["I love NLP"]
embeddings_df = nlu.load('en.embed.bert.small_L12_128').predict(text, output_level='token')
embeddings_df

Results

	token	en_embed_bert_small_L12_128_embeddings
		
	I	[-0.19200828671455383, -1.1298311948776245, -0...
	love	[-0.427543580532074, -1.2282991409301758, -0.2...
	NLP	[0.5775153636932373, 0.06353635340929031, -0.3...

Model Information

Model Name: small_bert_L12_128
Type: embeddings
Compatibility: Spark NLP 2.6.0+
License: Open Source
Edition: Official
Input Labels: [sentence, token]
Output Labels: [word_embeddings]
Language: [en]
Dimension: 128
Case sensitive: false

Data Source

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