|
gptkbp:instanceOf
|
large language model
|
|
gptkbp:architecture
|
gptkb:transformation
|
|
gptkbp:author
|
gptkb:Jacob_Devlin
gptkb:Kenton_Lee
gptkb:Ming-Wei_Chang
gptkb:Kristina_Toutanova
|
|
gptkbp:availableOn
|
gptkb:Hugging_Face_Model_Hub
gptkb:TensorFlow_Hub
|
|
gptkbp:bidirectional
|
true
|
|
gptkbp:citation
|
over 50,000
|
|
gptkbp:developedBy
|
gptkb:Google_AI
|
|
gptkbp:fineTunedWith
|
true
|
|
gptkbp:hiddenSize
|
1024
|
|
https://www.w3.org/2000/01/rdf-schema#label
|
BERT-large
|
|
gptkbp:impact
|
state-of-the-art on GLUE
state-of-the-art on SQuAD
|
|
gptkbp:input
|
gptkb:text
|
|
gptkbp:inputLengthLimit
|
512 tokens
|
|
gptkbp:introducedIn
|
2018
|
|
gptkbp:language
|
English
|
|
gptkbp:level
|
24
|
|
gptkbp:license
|
Apache 2.0
|
|
gptkbp:memiliki_tugas
|
question answering
sentiment analysis
named entity recognition
natural language inference
|
|
gptkbp:notableFor
|
deep bidirectional context
transformer-based pretraining
|
|
gptkbp:notablePublication
|
gptkb:BERT:_Pre-training_of_Deep_Bidirectional_Transformers_for_Language_Understanding
|
|
gptkbp:numberOfAttentionHeads
|
16
|
|
gptkbp:openSource
|
true
|
|
gptkbp:output
|
contextual embeddings
|
|
gptkbp:parameter
|
340 million
|
|
gptkbp:pretrainingObjective
|
masked language modeling
next sentence prediction
|
|
gptkbp:pretrainingTime
|
several days on 16 TPU chips
|
|
gptkbp:relatedTo
|
gptkb:BERT-base
gptkb:DistilBERT
gptkb:RoBERTa
|
|
gptkbp:tokenizerType
|
gptkb:WordPiece
|
|
gptkbp:trainer
|
gptkb:English_Wikipedia
gptkb:BooksCorpus
|
|
gptkbp:usedIn
|
information retrieval
search engines
chatbots
text classification
machine reading comprehension
|
|
gptkbp:usesSelfAttention
|
true
|
|
gptkbp:bfsParent
|
gptkb:TinyBERT
|
|
gptkbp:bfsLayer
|
6
|