No training configuration found in save file: the model was *not* compiled.

Thank you for the great work! But I got some problem.

I download the resnet50_coco pretrain model form here and loading it as prediction_model. but I get the problem as:

No training configuration found in save file

Although it said that is a safe warning, but I cant load the pretrain model with problem: IndexError: list index out of range. And I find that if I print print('step2',classification.shape) that will call twice of the filter_detections and the classification will get strange shape when it was called twice:

step1 (?, 80) step2 (?, 80) /home/XXX/workspace/tensorflow/lib/python3.5/site-packages/keras/models.py:255: UserWarning: No training configuration found in save file: the model was not compiled. Compile it manually. warnings.warn('No training configuration found in save file: ’ step1 (300,) step2 (300,) … File “/home/XXX/.local/lib/python3.5/site-packages/keras_retinanet/layers/filter_detections.py”, line 45, in filter_detections for c in range(int(classification.shape[1])): File “/home/XXX/workspace/tensorflow/lib/python3.5/site-packages/tensorflow/python/framework/tensor_shape.py”, line 521, in getitem return self._dims[key] IndexError: list index out of range

When I conver this pretrain model, this problem raise again. All of those problem will absent when I use a csv train model.

[hgaiser]

Did this fix your issue? If so, please close this issue.

[yelanyelan]

Please use --weight to load the pretrain model instead of --snapshot.