Marathi DistilBertForMaskedLM Cased model (from DarshanDeshpande)

Description

Pretrained DistilBertForMaskedLM model, adapted from Hugging Face and curated to provide scalability and production-readiness using Spark NLP. marathi-distilbert is a Marathi model originally trained by DarshanDeshpande.

Download Copy S3 URI

How to use

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

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

distilbert_loaded = DistilBertEmbeddings.pretrained("distilbert_embeddings_marathi","mr") \
    .setInputCols(["document", "token"]) \
    .setOutputCol("embeddings") \
    .setCaseSensitive(False)

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

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("document")
    .setOutputCol("token")

val distilbert_loaded = DistilBertEmbeddings.pretrained("distilbert_embeddings_marathi","mr")
    .setInputCols(Array("document", "token"))
    .setOutputCol("embeddings")
    .setCaseSensitive(false)

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

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

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

Model Information

Model Name: distilbert_embeddings_marathi
Compatibility: Spark NLP 4.2.4+
License: Open Source
Edition: Official
Input Labels: [sentence, token]
Output Labels: [embeddings]
Language: mr
Size: 247.8 MB
Case sensitive: false

References

  • https://huggingface.co/DarshanDeshpande/marathi-distilbert
  • https://github.com/DarshanDeshpande
  • https://www.linkedin.com/in/darshan-deshpande/
  • https://github.com/Baras64
  • http://​www.linkedin.com/in/harsh-abhi