Zobrazeno 1 - 10
of 359
pro vyhledávání: '"Balazy, P"'
The rapid expansion of large language models (LLMs) has underscored the need for parameter-efficient fine-tuning methods, with LoRA (Low-Rank Adaptation) emerging as a popular solution. Although LoRA reduces the number of trainable parameters, servin
Externí odkaz:
http://arxiv.org/abs/2405.17604
Autor:
Marta Ronowicz, Piotr Balazy, Maciej Chełchowski, Piotr Kuklinski, Weronika Patuła, Anna Sowa, Janne E. Søreide, Agata Weydmann-Zwolicka
Publikováno v:
Scientific Reports, Vol 14, Iss 1, Pp 1-13 (2024)
Abstract Benthic organisms typically possess a planktonic propagule stage in the form of larvae or spores, which enables them to spread over large distances before settlement, and promotes tight pelago-benthic coupling. However, factors driving dispe
Externí odkaz:
https://doaj.org/article/30cb463a229849eeaf0460bd631b519e
Autor:
Matteo Cecchetto, Agnès Dettai, Cyril Gallut, Matthias Obst, Piotr Kuklinski, Piotr Balazy, Maciej Chelchowski, Magdalena Małachowicz, Anita Poćwierz-Kotus, Małgorzata Zbawicka, Henning Reiss, Marc P. Eléaume, Gentile Francesco Ficetola, Christina Pavloudi, Katrina Exter, Diego Fontaneto, Stefano Schiaparelli
Publikováno v:
Nature Communications, Vol 15, Iss 1, Pp 1-15 (2024)
Abstract A pattern of increasing species richness from the poles to the equator is frequently observed in many animal taxa. Ecological limits, determined by the abiotic conditions and biotic interactions within an environment, are one of the major fa
Externí odkaz:
https://doaj.org/article/04529d12bfa84c6aa19c349a5db22c64
Nowadays artificial neural network models achieve remarkable results in many disciplines. Functions mapping the representation provided by the model to the probability distribution are the inseparable aspect of deep learning solutions. Although softm
Externí odkaz:
http://arxiv.org/abs/2304.05243
Autor:
Gaiński, Piotr, Bałazy, Klaudia
We propose a novel gradient-based attack against transformer-based language models that searches for an adversarial example in a continuous space of token probabilities. Our algorithm mitigates the gap between adversarial loss for continuous and disc
Externí odkaz:
http://arxiv.org/abs/2302.05120
Autor:
Banaei, Mohammadreza, Bałazy, Klaudia, Kasymov, Artur, Lebret, Rémi, Tabor, Jacek, Aberer, Karl
Recent transformer language models achieve outstanding results in many natural language processing (NLP) tasks. However, their enormous size often makes them impractical on memory-constrained devices, requiring practitioners to compress them to small
Externí odkaz:
http://arxiv.org/abs/2302.04045
Publikováno v:
Oceanologia, Vol 66, Iss 2, Pp 424-428 (2024)
The underwater meadows of the Puck Bay, once thriving with eelgrass Zostera marina, bladderwrack Fucus vesiculosus and black carrageen Furcellaria lumbricalis, experienced a decline in water quality during the 1960s and 1970s due to untreated sewage
Externí odkaz:
https://doaj.org/article/a392f45b62694211889dcf404912d84f
Autor:
Bernabé Moreno, Anna Sowa, Kamil Reginia, Piotr Balazy, Maciej Chelchowski, Marta Ronowicz, Piotr Kuklinski
Publikováno v:
Scientific Data, Vol 11, Iss 1, Pp 1-7 (2024)
Abstract Manifestations of climate change in the Arctic include an increase in water temperatures and massive loss of sea ice enabling more light penetration. Yet to understand tempo and scale of these parameters change over time, constant monitoring
Externí odkaz:
https://doaj.org/article/3d046a88bc2b4e569e11105a63e4b0c9
The adoption of Transformer-based models in natural language processing (NLP) has led to great success using a massive number of parameters. However, due to deployment constraints in edge devices, there has been a rising interest in the compression o
Externí odkaz:
http://arxiv.org/abs/2106.08181
Autor:
Wołczyk, Maciej, Wójcik, Bartosz, Bałazy, Klaudia, Podolak, Igor, Tabor, Jacek, Śmieja, Marek, Trzciński, Tomasz
The problem of reducing processing time of large deep learning models is a fundamental challenge in many real-world applications. Early exit methods strive towards this goal by attaching additional Internal Classifiers (ICs) to intermediate layers of
Externí odkaz:
http://arxiv.org/abs/2106.05409