|
gptkbp:instanceOf
|
gptkb:large_language_model
|
|
gptkbp:architecture
|
gptkb:transformation
|
|
gptkbp:author
|
gptkb:Ilya_Sutskever
gptkb:Alec_Radford
gptkb:Dario_Amodei
gptkb:David_Luan
gptkb:Rewon_Child
gptkb:Jeffrey_Wu
|
|
gptkbp:contextWindowSize
|
1024 tokens
|
|
gptkbp:developedBy
|
gptkb:OpenAI
|
|
gptkbp:input
|
gptkb:text
|
|
gptkbp:language
|
English
|
|
gptkbp:license
|
gptkb:OpenAI_License
|
|
gptkbp:memiliki_tugas
|
text generation
language modeling
|
|
gptkbp:notableFor
|
text completion
few-shot learning
multi-task learning
zero-shot learning
controversy over release
large-scale unsupervised learning
coherent long-form text generation
delayed full model release due to misuse concerns
demonstrated risks of large language models
influenced subsequent language models
no fine-tuning required for many tasks
|
|
gptkbp:notablePublication
|
gptkb:Language_Models_are_Unsupervised_Multitask_Learners
https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf
|
|
gptkbp:openSource
|
partially
|
|
gptkbp:output
|
gptkb:text
|
|
gptkbp:parameter
|
1.5 billion
|
|
gptkbp:predecessor
|
GPT (2018)
|
|
gptkbp:releaseYear
|
2019
|
|
gptkbp:size
|
1.5B
117M
345M
762M
|
|
gptkbp:successor
|
gptkb:GPT-3_(2020)
|
|
gptkbp:trainer
|
WebText dataset
|
|
gptkbp:trainingObjective
|
next word prediction
|
|
gptkbp:bfsParent
|
gptkb:OpenAI_models
|
|
gptkbp:bfsLayer
|
7
|
|
https://www.w3.org/2000/01/rdf-schema#label
|
GPT-2 (2019)
|