From c5424bf0b2f93124a3051f78babebf9191bbdeee Mon Sep 17 00:00:00 2001 From: skindhu Date: Thu, 7 Nov 2024 15:37:50 +0800 Subject: [PATCH] add fifth chapter --- cn-Book/5.在无标记数据集上进行预训练.md | 2 -- 1 file changed, 2 deletions(-) diff --git a/cn-Book/5.在无标记数据集上进行预训练.md b/cn-Book/5.在无标记数据集上进行预训练.md index 5a89050..f1f0910 100644 --- a/cn-Book/5.在无标记数据集上进行预训练.md +++ b/cn-Book/5.在无标记数据集上进行预训练.md @@ -278,8 +278,6 @@ tensor([ -9.5042, -10.3796, -11.3677, -11.4798, -9.7764, -12.2561]) > > 在计算交叉熵损失时,我们希望最大化模型分配给每个正确目标token的概率。交叉熵损失的数学公式为: > -> $$\text { Loss }=-\sum_{t=1}^{T} \ln P\left(y_{t} \mid x, \theta\right)$$ -> > 其中: > > + T 是序列长度