Guardado en:
Detalles Bibliográficos
Autores principales: Lomeli, Maria, Douze, Matthijs, Szilvasy, Gergely, Cabannes, Loic, Copet, Jade, Sukhbaatar, Sainbayar, Weston, Jason, Synnaeve, Gabriel, Mazaré, Pierre-Emmanuel, Jégou, Hervé
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2509.22358
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908730997604352
author Lomeli, Maria
Douze, Matthijs
Szilvasy, Gergely
Cabannes, Loic
Copet, Jade
Sukhbaatar, Sainbayar
Weston, Jason
Synnaeve, Gabriel
Mazaré, Pierre-Emmanuel
Jégou, Hervé
author_facet Lomeli, Maria
Douze, Matthijs
Szilvasy, Gergely
Cabannes, Loic
Copet, Jade
Sukhbaatar, Sainbayar
Weston, Jason
Synnaeve, Gabriel
Mazaré, Pierre-Emmanuel
Jégou, Hervé
contents We introduce stochastic activations. This novel strategy randomly selects between several non-linear functions in the feed-forward layer of a large language model. In particular, we choose between SILU or RELU depending on a Bernoulli draw. This strategy circumvents the optimization problem associated with RELU, namely, the constant shape for negative inputs that prevents the gradient flow. We leverage this strategy in two ways: (1) We use stochastic activations during pre-training and fine-tune the model with RELU, which is used at inference time to provide sparse latent vectors. This reduces the inference FLOPs and translates into a significant speedup on CPU and GPU. This leads to better results than training from scratch with the RELU activation function. (2) We evaluate stochastic activations for sequence generation. This strategy performs reasonably well: it has higher diversity and has only slightly inferior performance to the best deterministic non-linearity, SILU, combined with temperature sampling. This provides an alternative way to increase the diversity of generated text.
format Preprint
id arxiv_https___arxiv_org_abs_2509_22358
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Stochastic activations
Lomeli, Maria
Douze, Matthijs
Szilvasy, Gergely
Cabannes, Loic
Copet, Jade
Sukhbaatar, Sainbayar
Weston, Jason
Synnaeve, Gabriel
Mazaré, Pierre-Emmanuel
Jégou, Hervé
Machine Learning
Artificial Intelligence
We introduce stochastic activations. This novel strategy randomly selects between several non-linear functions in the feed-forward layer of a large language model. In particular, we choose between SILU or RELU depending on a Bernoulli draw. This strategy circumvents the optimization problem associated with RELU, namely, the constant shape for negative inputs that prevents the gradient flow. We leverage this strategy in two ways: (1) We use stochastic activations during pre-training and fine-tune the model with RELU, which is used at inference time to provide sparse latent vectors. This reduces the inference FLOPs and translates into a significant speedup on CPU and GPU. This leads to better results than training from scratch with the RELU activation function. (2) We evaluate stochastic activations for sequence generation. This strategy performs reasonably well: it has higher diversity and has only slightly inferior performance to the best deterministic non-linearity, SILU, combined with temperature sampling. This provides an alternative way to increase the diversity of generated text.
title Stochastic activations
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2509.22358