search

NVIDIA / gpu-rest-engine

A REST API for Caffe using Docker and Go
BSD 3-Clause "New" or "Revised" License
421 stars 94 forks source link

Problem when using model from DIGITS #11

Closed nicolasmetallo closed 7 years ago

nicolasmetallo commented 7 years ago

Hi,

I've trained my model under DIGITS 5 but when I try to make it work on GRE (following these steps) I keep receiving the same response, is it a bug because of the different versions of Caffe (DIGITS vs GRE)? Is there a workaround? Thanks!! Nicolas

[libprotobuf ERROR google/protobuf/text_format.cc:274] Error parsing text-format caffe.NetParameter: 53:18: Message type "caffe.BatchNormParameter" has no field named "scale_filler". F0422 07:17:48.987526 1 upgrade_proto.cpp:88] Check failed: ReadProtoFromTextFile(param_file, param) Failed to parse NetParameter file: /models/money_reader/deploy.prototxt

miraclewkf commented 7 years ago

have you known the reason?

flx42 commented 7 years ago

It could be because our fork nvidia/caffe has a different implementation of batch normalization, but I'm not an expert here. See https://github.com/NVIDIA/caffe/issues/276