All translations
Enter a message name below to show all available translations.
Found 3 translations.
| Name | Current message text |
|---|---|
| h English (en) | After release, {{Term|swish}} was added to mainstream frameworks (e.g. <code>tf.nn.swish</code>) and adopted in production architectures such as [[EfficientNet]]. The variant '''Hard {{Term|swish}}''' — a piecewise-linear approximation defined as <math>x \cdot \mathrm{ReLU6}(x + 3)/6</math> — was introduced in {{Term|mobilenet|MobileNetV3}} to recover {{Term|swish}}'s {{Term|accuracy}} gains while being cheap on mobile hardware. {{Term|gelu}} itself was later popularized by [[BERT (language model)|BERT]] and the {{Term|gpt}} family, where it became the default {{Term|activation function|activation}} in [[Transformer (machine learning model)|Transformer]] feed-forward blocks, vindicating the broader category that {{Term|swish}} helped make mainstream. |
| h Spanish (es) | Tras su publicación, {{Term|swish}} fue añadida a los marcos principales (p. ej. <code>tf.nn.swish</code>) y adoptada en arquitecturas de producción como [[EfficientNet]]. La variante '''Hard {{Term|swish}}''' — una aproximación lineal por tramos definida como <math>x \cdot \mathrm{ReLU6}(x + 3)/6</math> — se introdujo en {{Term|mobilenet|MobileNetV3}} para recuperar las ganancias de {{Term|accuracy|precisión}} de {{Term|swish}} mientras se mantiene económica en hardware móvil. La propia {{Term|gelu}} fue popularizada más tarde por [[BERT (language model)|BERT]] y la familia {{Term|gpt}}, donde se convirtió en la {{Term|activation function|función de activación}} predeterminada en los bloques feed-forward de los [[Transformer (machine learning model)|Transformer]], reivindicando la categoría más amplia que {{Term|swish}} ayudó a popularizar. |
| h Chinese (zh) | 发布之后,{{Term|swish}} 被加入主流框架(例如 <code>tf.nn.swish</code>),并被诸如 [[EfficientNet]] 等生产级架构采用。'''Hard {{Term|swish}}''' 变体 —— 定义为 <math>x \cdot \mathrm{ReLU6}(x + 3)/6</math> 的分段线性近似 —— 在 {{Term|mobilenet|MobileNetV3}} 中引入,以在保持移动端硬件低成本的同时恢复 {{Term|swish}} 的{{Term|accuracy|准确率}}增益。{{Term|gelu}} 本身后来由 [[BERT (language model)|BERT]] 和 {{Term|gpt}} 系列推广,在 [[Transformer (machine learning model)|Transformer]] 前馈块中成为默认的{{Term|activation function|激活函数}},验证了 {{Term|swish}} 帮助主流化的更广泛类别。 |