GPT2 text-to-text model, distilled version

Description

GPT-2 displays a broad set of capabilities, including the ability to generate conditional synthetic text samples of unprecedented quality, where the model is primed with an input and it generates a lengthy continuation. This is a distilled version which has fewer parameters and requires less computational resources to run.

Predicted Entities

Live Demo Open in Colab Download Copy S3 URI

How to use

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

gpt2 = GPT2Transformer.pretrained("gpt2_distilled") \\
.setInputCols(["documents"]) \\
.setMaxOutputLength(50) \\
.setOutputCol("generation")

pipeline = Pipeline().setStages([documentAssembler, gpt2])
data = spark.createDataFrame([["My name is Leonardo."]]).toDF("text")
result = pipeline.fit(data).transform(data)
result.select("summaries.generation").show(truncate=False)

val documentAssembler = new DocumentAssembler()
.setInputCol("text")
.setOutputCol("documents")

val gpt2 = GPT2Transformer.pretrained("gpt2_distilled")
.setInputCols(Array("documents"))
.setMinOutputLength(10)
.setMaxOutputLength(50)
.setDoSample(false)
.setTopK(50)
.setNoRepeatNgramSize(3)
.setOutputCol("generation")

val pipeline = new Pipeline().setStages(Array(documentAssembler, gpt2))
val data = Seq("My name is Leonardo.").toDF("text")
val result = pipeline.fit(data).transform(data)
results.select("generation.result").show(truncate = false)

import nlu
nlu.load("en.gpt2.distilled").predict("""My name is Leonardo.""")

Model Information

Data Source

OpenAI WebText - a corpus created by scraping web pages with emphasis on document quality. It consists of over 8 million documents for a total of 40 GB of text. All Wikipedia documents were removed from WebText.