detecto
detecto copied to clipboard
Plot Training Loss
Describe the feature you'd like I would like to see the training loss not only the validation loss
Describe the use cases of the feature To better detect cases of under/overfitting
Additional context
Thanks!
Suggestion noted. In the meantime, a workaround could be to copy and modify the source code for the fit
method to return both the training and validation losses.
It would be really nice to have a progress bar as well.
@cagbal just want to confirm: have you been using the verbose=True
flag when training?
Exactly. I trained the system with 10K images with bbs. It was hard to understand how long would one epoch take. A progress bar would help a lot.
@alankbi Would it be possible to link it to tensorboard?
Maybe this tutorial can help? With Detecto's Model class, you can call the get_internal_model
method to get the actual PyTorch model, which you can then pass to the SummaryWriter as detailed in that tutorial.
Update: a progress bar has been added in v1.1.2 :)