Caffe cont'd training doesn't find custom Python layer

[Alex Ter-Sarkisov]

I'm continuing from snapshot: 

"/caffe/build/tools/caffe" train -solver solver.prototxt -snapshot snapshot/train_iter_1.solverstate

and the error is 

I1204 12:17:23.885371 22143 layer_factory.hpp:77] Creating layer maskextractor
ImportError: No module named mask_extractor
terminate called after throwing an instance of 'boost::python::error_already_set'
*** Aborted at 1512389843 (unix time) try "date -d @1512389843" if you are using GNU date ***
PC: @     0x7f027d61bc37 (unknown)
*** SIGABRT (@0x5296f0000567f) received by PID 22143 (TID 0x7f027f415780) from PID 22143; stack trace: ***

I've written Python script with imports to sys.path: 

import sys
newmodule_dir = "../lib/mask_extractor"
sys.path.insert(0, newmodule_dir)
import mask_extractor
call('/home/ICTDOMAIN/453615/Downloads/fcn.berkeleyvision.org/voc-fcn8s/run_ft.sh', shell=True)

When I run the code with solver instead of cont'd training, it works fine. Other errors pop up, but it's a different story. 

So what am I doing wrong here?