Malay DistilBERT Embeddings (from w11wo)

Description

Pretrained DistilBERT Embeddings model, uploaded to Hugging Face, adapted and imported into Spark NLP. malaysian-distilbert-small is a Malay model orginally trained by w11wo.

Download

How to use

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

tokenizer = Tokenizer() \
    .setInputCols("document") \
    .setOutputCol("token")
  
embeddings = DistilBertEmbeddings.pretrained("distilbert_embeddings_malaysian_distilbert_small","ms") \
    .setInputCols(["document", "token"]) \
    .setOutputCol("embeddings")
    
pipeline = Pipeline(stages=[documentAssembler, tokenizer, embeddings])

data = spark.createDataFrame([["Saya suka 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 = DistilBertEmbeddings.pretrained("distilbert_embeddings_malaysian_distilbert_small","ms") 
    .setInputCols(Array("document", "token")) 
    .setOutputCol("embeddings")

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

val data = Seq("Saya suka Spark NLP").toDF("text")

val result = pipeline.fit(data).transform(data)
import nlu
nlu.load("ms.embed.distilbert").predict("""Saya suka Spark NLP""")

Model Information

Model Name: distilbert_embeddings_malaysian_distilbert_small
Compatibility: Spark NLP 3.4.2+
License: Open Source
Edition: Official
Input Labels: [sentence, token]
Output Labels: [bert]
Language: ms
Size: 248.4 MB
Case sensitive: true

References

  • https://huggingface.co/w11wo/malaysian-distilbert-small
  • https://arxiv.org/abs/1910.01108
  • https://github.com/sgugger
  • https://github.com/piegu/fastai-projects/blob/master/finetuning-English-GPT2-any-language-Portuguese-HuggingFace-fastaiv2.ipynb
  • https://w11wo.github.io/