Is the initial learning rate 0.001 or 0.00025?

msracver / Flow-Guided-Feature-Aggregation

Flow-Guided Feature Aggregation for Video Object Detection

MIT License

723 stars 190 forks source link

Is the initial learning rate 0.001 or 0.00025? #41

Open kindloaf opened 5 years ago

kindloaf commented 5 years ago

In the paper, the learning rate is 0.001. However, in the source code, the learning rate is 0.00025, as set in the yaml file. Why there is such a difference?