ruishu / micro-projects Goto Github PK

Hi, RuiShu,
I tried to run the code tf-ema. It showed a bug like this:

{'design': 'v1', 'logdir': 'log', 'lr': 0.001, 'run': 999, 'src': 'cifar10'}
Model name: model=classifier_src=cifar10_design=v1_run=0999
Traceback (most recent call last):
File "run_classifier.py", line 34, in
M = classifier()
File "/home/leifan/micro-projects-master/tf-ema/codebase/models/classifier.py", line 32, in classifier
src_y = des.classifier(T.src_x, T.phase, internal_update=True)
File "/home/leifan/micro-projects-master/tf-ema/codebase/models/designs/v1.py", line 39, in classifier
x = f(x, *f_args, **f_kwargs)
File "/home/leifan/tensorflow/_python_build/tensorflow/contrib/framework/python/ops/arg_scope.py", line 177, in func_with_args
return func(*args, **current_args)
File "/home/leifan/.local/lib/python2.7/site-packages/tensorbayes/layers/simple.py", line 66, in conv2d
initializer=tf.zeros_initializer)
File "/home/leifan/tensorflow/_python_build/tensorflow/python/ops/variable_scope.py", line 987, in get_variable
custom_getter=custom_getter)
File "/home/leifan/tensorflow/_python_build/tensorflow/python/ops/variable_scope.py", line 889, in get_variable
custom_getter=custom_getter)
File "/home/leifan/tensorflow/_python_build/tensorflow/python/ops/variable_scope.py", line 347, in get_variable
validate_shape=validate_shape)
File "/home/leifan/tensorflow/_python_build/tensorflow/python/ops/variable_scope.py", line 332, in _true_getter
caching_device=caching_device, validate_shape=validate_shape)
File "/home/leifan/tensorflow/_python_build/tensorflow/python/ops/variable_scope.py", line 683, in _get_single_variable
validate_shape=validate_shape)
File "/home/leifan/tensorflow/_python_build/tensorflow/python/ops/variables.py", line 226, in init
expected_shape=expected_shape)
File "/home/leifan/tensorflow/_python_build/tensorflow/python/ops/variables.py", line 303, in _init_from_args
initial_value(), name="initial_value", dtype=dtype)
File "/home/leifan/tensorflow/_python_build/tensorflow/python/ops/variable_scope.py", line 672, in
shape.as_list(), dtype=dtype, partition_info=partition_info)
TypeError: zeros_initializer() got multiple values for keyword argument 'dtype'

I saw your blog post here that was super informative:
http://ruishu.io/2017/11/22/ema/

I'm trying to reproduce EMA for YOLOv3 in PyTorch and am wondering if I should apply it after every batch or every optimizer update? Thank you!
https://github.com/ultralytics/yolov3