T5

E435867

T5 is a Transformer-based text-to-text language model developed by Google that treats every NLP task as converting input text to output text.

All labels observed (5)

Label Occurrences
T5 canonical 5
T5-11B 1
T5-3B 1

How this entity was disambiguated

Statements (47)

Predicate Object
instanceOf Transformer-based model
neural language model
text-to-text model
abbreviationOf Text-to-Text Transfer Transformer
basedOn Transformer architecture
citationVenue Journal of Machine Learning Research
linked to: JMLR
developer Google
Google Research
frameworkImplementation JAX
TensorFlow
hasVariant T5-11B
linked to: T5

T5-3B
linked to: T5

T5-Base
linked to: T5

T5-Large
T5-Small
linked to: T5
inputFormat text
inspired unified text-to-text framework for NLP
introducedBy Adam Roberts
Colin Raffel
Katherine Lee
Michael Matena
Noam Shazeer
Peter J. Liu
Sharan Narang
Wei Li
Yanqi Zhou
language English
license Apache License 2.0
modelType encoder-decoder
openSource true
optimizationGoal transfer learning across NLP tasks
pretrainingObjective denoising autoencoding
span corruption
publicationTitle Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
publicationYear 2019
releasedYear 2019
supportsFineTuning true
supportsTask machine translation
natural language inference
question answering
summarization
text classification
text generation
trainingData Colossal Clean Crawled Corpus
treatsEveryNLPTasAs text-to-text problem
usesDecoder Transformer decoder
usesEncoder Transformer encoder

How these facts were elicited

Referenced by (9)

Full triples — surface form annotated when it differs from this entity's canonical label.

T5 hasVariant T5-Small
linked to: T5
T5 hasVariant T5-Base
linked to: T5
T5 hasVariant T5-3B
linked to: T5
T5 hasVariant T5-11B
linked to: T5