- Generative Adversarial Networks Projects
- Kailash Ahirwar
- 127字
- 2021-07-02 13:38:53
Visualizing losses
To visualize the losses for the training, start the tensorboard server, as follows:
tensorboard --logdir=logs
Now, open localhost:6006 in your browser. The SCALARS section of TensorBoard contains plots for both losses:

Loss plot for the generator network

Loss plot for the discriminator network
These plots will help you to decide whether to continue or stop the training. If the losses are not decreasing anymore, you can stop the training, as there is no chance of improvement. If the losses keep increasing, you must stop the training. Play with the hyperparameters and select a set of hyperparameters that you think might provide better results. If the losses are decreasing gradually, keep training the model.
推薦閱讀
- Oracle SOA Governance 11g Implementation
- 高性能混合信號(hào)ARM:ADuC7xxx原理與應(yīng)用開(kāi)發(fā)
- R Machine Learning By Example
- 輕松學(xué)Java
- 流處理器研究與設(shè)計(jì)
- 嵌入式Linux上的C語(yǔ)言編程實(shí)踐
- Expert AWS Development
- 深度學(xué)習(xí)中的圖像分類與對(duì)抗技術(shù)
- 數(shù)據(jù)挖掘方法及天體光譜挖掘技術(shù)
- AWS Certified SysOps Administrator:Associate Guide
- 我也能做CTO之程序員職業(yè)規(guī)劃
- RedHat Linux用戶基礎(chǔ)
- 筆記本電腦維修90個(gè)精選實(shí)例
- The DevOps 2.1 Toolkit:Docker Swarm
- FANUC工業(yè)機(jī)器人虛擬仿真教程