Training Question

[HoHoHO1999]

Train Epoch: 4 [49920/141886 (35%)]     Loss: 0.326567
grad_norm: 4.0071 iteration: 59448
loss_radarClass:0.134  loss_radarOffset:0.135  loss_radarDepthOffset:0.057  loss:0.327  
Train Epoch: 4 [49960/141886 (35%)]     Loss: 0.806502
grad_norm: 5.5651 iteration: 59453
loss_radarClass:0.214  loss_radarOffset:0.465  loss_radarDepthOffset:0.128  loss:0.807  
Train Epoch: 4 [50000/141886 (35%)]     Loss: 0.517228
grad_norm: 4.8307 iteration: 59458
loss_radarClass:0.126  loss_radarOffset:0.275  loss_radarDepthOffset:0.117  loss:0.517  
Train Epoch: 4 [50040/141886 (35%)]     Loss: 0.772954
grad_norm: 4.2956 iteration: 59463
loss_radarClass:0.192  loss_radarOffset:0.481  loss_radarDepthOffset:0.100  loss:0.773  
Train Epoch: 4 [50080/141886 (35%)]     Loss: 0.901844
grad_norm: 5.4652 iteration: 59468
loss_radarClass:0.153  loss_radarOffset:0.586  loss_radarDepthOffset:0.163  loss:0.902  
Train Epoch: 4 [50120/141886 (35%)]     Loss: 0.904489
grad_norm: 6.5861 iteration: 59473
loss_radarClass:0.184  loss_radarOffset:0.549  loss_radarDepthOffset:0.171  loss:0.904  
Train Epoch: 4 [50160/141886 (35%)]     Loss: 0.746246
grad_norm: 5.4594 iteration: 59478
loss_radarClass:0.185  loss_radarOffset:0.452  loss_radarDepthOffset:0.109  loss:0.746  

Hello @longyunf , this is a part of my training log, may I ask what all of these loss mean?

• Loss
• grad_norm
• iteration
• loss_radarClass
• loss_radarOffset
• loss_radarDepthOffset
• loss

and some times the value of the loss is higher and lower, is this normal? Thanks