: Vocabulario de PLN: fine-tuning

... > artificial intelligence > machine learning > machine learning approach > transfer learning > fine-tuning

fine-tuning

Approach to transfer learning in which the weights of a pre-trained model are trained on new data. ( https://d2l.ai/chapter_computer-vision/fine-tuning.html)

Fine tuning is a popular domain adaptation method which trains a neural language model in two phases first maximizing the likelihood of the generic set D (pre-training) before optimizing the likelihood of the target domain set T (fine-tuning). (Grangier & Iter, 2022)

Furthermore we show that fine-tuning the connector layers frequently enhances performance within MLLMs. (Zhou, He, Ke, Zhu, Gutierrez Basulto & Pan, 2024)
In this paper we presented an approach for predicting empathic concern and personal distress by fine- tuning a pre-trained language model using parameter sharing. (Butala, Singh, Kumar & Shrivastava, 2021)
The fine-tuning is done with mini-batch gradient descent for the classification layer and a number of self-attention layers in the backbone. (Holur, Wang, Shahsavari, Tangherlini & Roychowdhury, 2022)
The pretraining and fine tuning described in Section 4.1 took less than 20 hours using a single GPU. (Ciosici, Cummings, DeHaven, Hedges, Kankanampati, Lee, Weischedel & Freedman, 2021)
We show that the pre-finetuned models consistently require less data for fine-tuning. (Aghajanyan, Gupta, Shrivastava, Chen, Zettlemoyer & Gupta, 2021)

http://data.loterre.fr/ark:/67375/8LP-X09CB8JF-R

RDF/XML TURTLE JSON-LD última modificación 30/9/24

Loterre