How do you resume training?

[1yian]

What specific flags should I pass into train.py in order to resume my training checkpoint? I'm not sure if --transfer none, fine_tune, freeze, darknet fit the criteria to resume as they all involve freezing or getting rid of some parts of the weights.

[Hugh0120]

I simply replaced that part of code with
model.load_weights("/path/to/your/previous/ckpt")
to resume training

[ilyes-hamrouni]

I simply replaced that part of code with
model.load_weights("/path/to/your/previous/ckpt")
to resume training

How can do it for training from scratch there is no model.load in the no transfer.

[studjon]

I simply replaced that part of code with
model.load_weights("/path/to/your/previous/ckpt")
to resume training

How can do it for training from scratch there is no model.load in the no transfer.

You can simply cut the line "model.load_weights(FLAGS.weights)" and paste it before the "if transfer != none". It will then load the weights you specified. This weight flag then should contain the path to your previously trained weights, something like "./dirPath/yolov3_train_100.tf". Hope this is still of use, if not, maybe for future users having problems with this part.

[TheClassyPenguin]

Have any of you successfully resumed fine-tuning training of darknet with custom classes?

I tried the method mentioned above but I got:

ValueError: Shapes (18,) and (255,) are incompatible

This is the same error I'd get if I didn't specify the --weight_classes parameter.