I have a trained model with some custom classes but I need to fine tune it adding new tagged examples to the training set.
I've already updated the tfrecords but I can't resume the training job after it is canceled and I also can´t start a new training job with the old job_id.
All I could do is starting a new training job from zero with the updated tfrecords but it trains from the resnet checkpoint instead of my pretrained checkpoint.
So, is there a way to resume a training job after modifying the train.tfrecords?
I have a trained model with some custom classes but I need to fine tune it adding new tagged examples to the training set. I've already updated the tfrecords but I can't resume the training job after it is canceled and I also can´t start a new training job with the old job_id.
All I could do is starting a new training job from zero with the updated tfrecords but it trains from the resnet checkpoint instead of my pretrained checkpoint.
So, is there a way to resume a training job after modifying the train.tfrecords?