Resume training from saved model.step in detectnet_v2

Accelerated Computing Intelligent Video Analytics TAO Toolkit
#1

Hi,

How to resume training from saved model.step-<step_no>.tlt (model.step-1498.tlt) in detectnet_v2.

My training always starts with step-0.

tlt-train detectnet_v2 -r /op1/ -e /config  -k $key

model_config {

 pretrained_model_file: "(path-to)/model.step-1498.tlt"

.....
}
#2

Hi sathiez,
Thank you so much for using TLT! Unfortunately,for this release,we don’t support resuming training from a checkpoint.We will look into adding this as a feature for our next release.

#3

Do you have any update, where we can resume our model and start training?
Thank You,

#4

Hi ajayskabadi2012,
It is a new feature which is designed underway.
For current release,we don’t support resuming training from a checkpoint.

#5