Hello,
I am trying to convert a pretrained TensorFlow InceptionV3 to dlc. The net strucutre this https://github.com/tensorflow/models/blob/master/research/slim/nets/ince...
But I see an error "RuntimeWarning: error_code=1004; error_message=Layer parameters combination is invalid. Layer InceptionV3/InceptionV3/Mixed_7a/concat: input InceptionV3/InceptionV3/Mixed_7a/Branch_2/MaxPool_1a_3x3/MaxPool:0 has size 4 along axis 0, should match output dim (3); error_component=Model Validation; line_no=516; thread_id=140619388966656
descriptor.axis)"
I don't understand what the converter complaining b/c there is nothing wrong with net atchitecture otherwise it would have complained by TensorFLow when I trained. I can train this model without problem for sure.
And also even more strange, in https://github.com/tensorflow/models/blob/master/research/slim/nets/ince..., Mixed_6a has the same block deisgn as MIxed_7a and I don't see the converter complains Mixed_6a
Can anyone help?