Overfitting models by using tensorboard

Accelerated Computing Intelligent Video Analytics TAO Toolkit
5

Several comments here.

  1. For detectnet_v2, during training, there should be printing log for AP, mAP or loss. Please double check if you enable below. It will print AP or mAP info every 10 epochs.
validation_period_during_training: 10 
first_validation_epoch: 1

  1. For tensorboard, please refer to Output file when training detectnet 2

  2. For retinanet, there should be training logs. Can you elaborate your issue, why did you say there is no training logs?