Results 21 to 30 of about 14,768 (170)
CoRRLoRA (Low-Rank Adaptation) has emerged as a preferred method for efficiently adapting Large Language Models (LLMs) with remarkable simplicity and efficacy. This note extends the original LoRA paper by offering new perspectives that were not initially discussed and presents a series of insights for deploying LoRA at scale.
Vlad Fomenko +4 more
openaire +2 more sources
Dual LoRA: Enhancing LoRA with Magnitude and Direction Updates
CoRRLow-rank adaptation (LoRA) is one of the most popular methods among parameter-efficient fine-tuning (PEFT) methods to adapt pre-trained large language models (LLMs) to specific downstream tasks. However, the model trained based on LoRA often has an unsatisfactory performance due to its low-rank assumption.
Yixing Xu +6 more
openaire +2 more sources
Run LoRA Run: Faster and Lighter LoRA Implementations
Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track)LoRA is a technique that reduces the number of trainable parameters in a neural network by introducing low-rank adapters to linear layers. This technique is used both for fine-tuning and full training of large language models. This paper presents the RunLoRA framework for efficient implementations of LoRA that significantly improves the speed of neural
Daria Cherniuk +2 more
openaire +2 more sources
FedRot-LoRA: Mitigating Rotational Misalignment in Federated LoRA
CoRRFederated LoRA provides a communication-efficient mechanism for fine-tuning large language models on decentralized data. In practice, however, a discrepancy between the factor-wise averaging used to preserve low rank and the mathematically correct aggregation of local updates can cause significant aggregation error and unstable training.
Haoran Zhang +3 more
openaire +2 more sources
Tied-LoRA: Enhancing parameter efficiency of LoRA with Weight Tying
Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)8 pages 4 ...
Adithya Renduchintala +2 more
openaire +2 more sources
Exploring the Security Vulnerabilities of LoRa [PDF]
2017 3rd IEEE International Conference on Cybernetics (CYBCONF), 2017 Internet-of-Things (IoT) deployments increasingly incorporate long range communication technologies. To support this transition, wide area IoT deployments are employing LoRa as their communication technology of choice due to its low power consumption and long range.
Aras, Emekcan +3 more
openaire +2 more sources
Clinical epidemiological, endoscopic and histologic aspects in elderly with esophagus cancer
Medisan, 2022 Introduction: The esophagus cancer is one of the most invasive neoplasms and it means a less survival, because it is generally diagnosed in a late way, mainly in people over 60 years.
Yurislena Monier Martínez +4 more
doaj
Conocimientos vigentes acerca del cáncer de colon recurrente
Medisan, 2015 Aproximadamente hasta 30 % de todos los pacientes con cáncer de colon, intervenidos con intención curativa, presentarán enfermedad recurrente, la cual puede provocar la muerte o motivar nuevos tratamientos con alto grado de morbilidad, sin que se haya ...
Zenén Rodríguez Fernández +4 more
doaj
Medisan, 2014 Se realizó un estudio descriptivo, observacional y longitudinal de 478 pacientes egresados con el diagnóstico histopatológico de apendicitis aguda, que habían sido operados mediante apendicectomía videolaparoscópica, en el Servicio de Cirugía General del
Benjamín Deulofeu Betancourt +4 more
doaj
Factores predictivos de mortalidad en pacientes con oclusión intestinal mecánica
Medisan, 2015 Se realizó un estudio longitudinal, prospectivo, de cohorte, de 191 pacientes con oclusión intestinal mecánica, operados en el Servicio de Cirugía del Hospital Provincial Docente Clinicoquirúrgico "Saturnino Lora Torres" de Santiago de Cuba, desde enero ...
Erian Jesús Domínguez González +3 more
doaj