I am trying to train the TLT Detectnet_V2 on my custom data set. I have 1000 images in the data set.
When I try to train it I get this error
File "./detectnet_v2/cost_function/cost_auto_weight_hook.py", line 26, in build_cost_auto_weight_hook ValueError: steps_per_epoch must be > 0
The TLT Detecnet_V2 trains just fine on data sets that are at least 7000 images.
I have read the documentation and it dose not explain how to change the “steps_per_epoch”
I have tried tweeking on the “training_config” of the spec file but it doesnt make a difference.