English RoBERTa Embeddings (Mixed sampling strategy)

Description

Pretrained RoBERTa Embeddings model, uploaded to Hugging Face, adapted and imported into Spark NLP. distilroberta-base-climate-d-s is a English model orginally trained by climatebert.

Sampling strategy ds:As expressed in the author’s paper here, ds is “div select + sim select”, meaning 70% of the biggest composite scaled score diverse+sim was used, discarding the rest.

Download Copy S3 URI

How to use

documentAssembler = DocumentAssembler() \
.setInputCol("text") \
.setOutputCol("document")

tokenizer = Tokenizer() \
.setInputCols("document") \
.setOutputCol("token")

embeddings = RoBertaEmbeddings.pretrained("roberta_embeddings_distilroberta_base_climate_d_s","en") \
.setInputCols(["document", "token"]) \
.setOutputCol("embeddings")

pipeline = Pipeline(stages=[documentAssembler, tokenizer, embeddings])

data = spark.createDataFrame([["I love Spark NLP"]]).toDF("text")

result = pipeline.fit(data).transform(data)
val documentAssembler = new DocumentAssembler() 
.setInputCol("text") 
.setOutputCol("document")

val tokenizer = new Tokenizer() 
.setInputCols(Array("document"))
.setOutputCol("token")

val embeddings = RoBertaEmbeddings.pretrained("roberta_embeddings_distilroberta_base_climate_d_s","en") 
.setInputCols(Array("document", "token")) 
.setOutputCol("embeddings")

val pipeline = new Pipeline().setStages(Array(documentAssembler, tokenizer, embeddings))

val data = Seq("I love Spark NLP").toDF("text")

val result = pipeline.fit(data).transform(data)
import nlu
nlu.load("en.embed.distilroberta_base_climate_d_s").predict("""I love Spark NLP""")

Model Information

Model Name: roberta_embeddings_distilroberta_base_climate_d_s
Compatibility: Spark NLP 3.4.2+
License: Open Source
Edition: Official
Input Labels: [sentence, token]
Output Labels: [bert]
Language: en
Size: 310.0 MB
Case sensitive: true

References

  • https://huggingface.co/climatebert/distilroberta-base-climate-d-s
  • https://arxiv.org/abs/2110.12010