Vorrichtung und Computerimplementiertes Verfahren zur Anpassung eines Modells an Aufgaben

Details

Veröffentlichungs-Nr.

EP4718283

Anmeldetag

25. September 2024

Veröffentlichung

1. April 2026

Rechtsraum

EP

IPC

(GOOGLE LLC et al.)(ZHAO ZHE et al.)

Abstract

Device, data structure, and method for adapting model to tasks, wherein the method comprises providing the model, wherein the model comprises an in particular linear layer for mapping a multidimensional input of the layer depending on weights to a multidimensional output of the layer, wherein the model comprises experts (208, 210) and a router gate (204) for adapting the model to different tasks, wherein the method comprises providing the input to the router gate (204), determining an output of the experts (208, 210) depending on an output of the router gate (204) in response to the input, modifying the model depending on the output of the experts (208, 210), mapping the input with the modified layer to the output of the model, training a first expert (208) of the experts with a first training method depending on the output of the model, and training a second expert (210) of the experts with a second training method depending on the output of the model, and maintaining the weights and the second expert (210) unchanged in the training with the first training method, and maintaining the weights and the first expert (208) unchanged in the training with the second training method.

Anmelder

Firma

Robert Bosch GmbH

Land

🇩🇪 Deutschland