Google Colab 中使用 tensorboard 绘制训练损失 loss 曲线

首先，在代码中使用 tensorboard 保存训练过程的 loss（确保已经安装了 tensorboard），可参考https://blog.csdn.net/haoshan4783/article/details/89970227

训练完毕后，使用以下代码绘制 loss 曲线：

%load_ext tensorboard
%tensorboard --logdir /your/content/directory

如果出现以下报错：

ERROR: Timed out waiting for TensorBoard to start. It may still be running as pid 4077.

则在上面的代码改为：

!pip uninstall -y tensorboard tb-nightly
!pip install -U tb-nightly
%load_ext tensorboard
%tensorboard --logdir /your/content/directory

如果出现以下报错：

TensorBoard attempted to bind to port 6006, but it was already in use

则需要手动指定端口（多尝试几个不同的端口）：

!pip uninstall -y tensorboard tb-nightly
!pip install -U tb-nightly
%load_ext tensorboard
%tensorboard --logdir /your/content/directory --port 8800