In [3]:

    

import numpy as np
import pandas as pd
import random
import fnmatch
import os
import h5py

import tensorflow as tf
print('TF file:',tf.__file__)

path = '/home/jorge/data/kaggle/statefarm/'


    

    




('TF file:', '/usr/local/lib/python2.7/dist-packages/tensorflow/__init__.pyc')

In [4]:

    

#Use the pretrained model
path_model = '/home/jorge/data/pretrained_models/tensorflow/inception/'
#path_model = '/Users/jorge/tensorflow/tensorflow/models/image/imagenet'

#Create graph
with tf.gfile.FastGFile(os.path.join(path_model, 'classify_image_graph_def.pb'), 'rb') as f:
    graph_def = tf.GraphDef()
    graph_def.ParseFromString(f.read())
    _ = tf.import_graph_def(graph_def, name='')


    

In [50]:

    

def evaluate_inception(batch_img):
    
    mixed2_conv = np.empty([len(batch_img), 35, 35, 288], dtype=np.float32)
        
    mixed8_conv = np.empty([len(batch_img), 8, 8, 1280])
    mixed8_tensor_values = np.empty([len(batch_img), 1280])
    
    mixed9_conv = np.empty([len(batch_img), 8, 8, 2048])
    mixed9_tensor_values = np.empty([len(batch_img), 2048])
    
    mixed10_conv = np.empty([len(batch_img), 8, 8, 2048])
    incepcion_pool3_values = np.empty([len(batch_img), 2048])

    predictions = np.empty([len(batch_img),1008])
    image_list = []

    #Extract intermediate data from the jpg file
    with tf.Session() as sess:
        #softmax_tensor = sess.graph.get_tensor_by_name('final_result:0')
        softmax_tensor = sess.graph.get_tensor_by_name('softmax:0')

        mixed2_tensor = sess.graph.get_tensor_by_name('mixed_2/join:0')

        mixed8_tensor = sess.graph.get_tensor_by_name('mixed_8/join:0')

        mixed9_tensor = sess.graph.get_tensor_by_name('mixed_9/join:0')

        mixed10_tensor = sess.graph.get_tensor_by_name('mixed_10/join:0')
        pool3_tensor = sess.graph.get_tensor_by_name('pool_3/_reshape:0')
        #print pool3_tensor.get_shape() 
        #print pool3_tensor[:,0].eval()

        for i, img in enumerate(batch_img):
            image_list += [img.split('/')[-1]]

            image_data = tf.gfile.FastGFile(img,'rb').read()
            
            mixed2_tensor_value = sess.run(mixed2_tensor, {'DecodeJpeg/contents:0': image_data})
            mixed2_conv[i] = mixed2_tensor_value[0,:,:,:]
            
            mixed8_tensor_value = sess.run(mixed8_tensor, {'mixed_2/join:0': mixed2_tensor_value})
            mixed8_conv[i] = mixed8_tensor_value[0,:,:,:]
            mixed8_tensor_values[i] = np.max(np.reshape(mixed8_tensor_value[0,:,:,:],[8*8,1280]),axis=0)

            mixed9_tensor_value = sess.run(mixed9_tensor, {'mixed_8/join:0': mixed8_tensor_value})
            mixed9_conv[i] = mixed9_tensor_value[0,:,:,:]
            mixed9_tensor_values[i] = np.max(np.reshape(mixed9_tensor_value[0,:,:,:],[8*8,2048]),axis=0)

            mixed10_tensor_value = sess.run(mixed10_tensor, {'mixed_9/join:0': mixed9_tensor_value})
            mixed10_conv[i] = mixed10_tensor_value[0,:,:,:]
            
            incepcion_pool3_value = sess.run(pool3_tensor,{'mixed_9/join:0': mixed9_tensor_value})
            incepcion_pool3_values[i] = incepcion_pool3_value[0,:]

            predictions[i] = sess.run(softmax_tensor,{'pool_3/_reshape:0': incepcion_pool3_value})[0,:]
            if i % 500 == 0:
                print('Processed', i, 'images from ',len(batch_img))
    return image_list, mixed2_conv, mixed8_conv, mixed8_tensor_values, mixed9_conv, mixed9_tensor_values, mixed10_conv, incepcion_pool3_values, predictions


    

In [51]:

    

#Test
image_list, conv2, conv8, layer8, conv9, layer9, conv10, layer10, predictions = evaluate_inception(['/home/jorge/data/kaggle/statefarm//sources/imgs/train/c2/img_48527.jpg',
                                                          '/home/jorge/data/kaggle/statefarm//sources/imgs/train/c2/img_2782.jpg'])
print('layer8: ', layer8[0])
print('layer9: ', layer9[0])
print('layer10: ', layer10[0])
print('predictions: ', predictions[0])
print('Image list: ', image_list)


    

    




('Processed', 0, 'images from ', 2)
('layer8: ', array([ 1.35979533,  2.66526389,  2.13184237, ...,  1.7782408 ,
        1.33178174,  1.58746636]))
('layer9: ', array([ 0.43170908,  1.72573614,  1.11879694, ...,  0.24815984,
        0.18799247,  0.58042067]))
('layer10: ', array([ 0.31927314,  0.79435837,  0.52067602, ...,  0.55574703,
        0.06320436,  0.27538911]))
('predictions: ', array([  7.62325217e-05,   3.40247243e-05,   1.40678909e-04, ...,
         7.62327400e-05,   7.62314376e-05,   7.62314376e-05]))
('Image list: ', ['img_48527.jpg', 'img_2782.jpg'])

In [15]:

    

def evaluate_inception_hdf5(batch_img, hdf5_file, batch_size=64):
    
    if os.path.isfile(hdf5_file):
        os.remove(hdf5_file)
        
    with h5py.File(hdf5_file, "a") as hdf5_f:
        
        mixed2_conv=[]
        mixed4_conv=[]
        mixed6_conv=[]
        mixed8_conv=[]
        mixed9_conv=[]
        mixed10_conv=[]
        predictions=[]

        image_list = []

        #Extract intermediate data from the jpg file
        with tf.Session() as sess:
            
            softmax_tensor = sess.graph.get_tensor_by_name('softmax:0')

            mixed2_tensor = sess.graph.get_tensor_by_name('mixed_2/join:0')

            mixed4_tensor = sess.graph.get_tensor_by_name('mixed_4/join:0')
            
            mixed6_tensor = sess.graph.get_tensor_by_name('mixed_6/join:0')

            mixed8_tensor = sess.graph.get_tensor_by_name('mixed_8/join:0')

            mixed9_tensor = sess.graph.get_tensor_by_name('mixed_9/join:0')

            mixed10_tensor = sess.graph.get_tensor_by_name('mixed_10/join:0')
            
            batch_step=0
            for i, img in enumerate(batch_img):
                
                if (i>0 and i%batch_size==0):
                    #Create conv2
                    hdf5_f.create_dataset("conv2_"+str(batch_step), data = np.array(mixed2_conv, dtype=np.float32))
                    mixed2_conv=[]

                    #Create conv4
                    hdf5_f.create_dataset("conv4_"+str(batch_step), data = np.array(mixed4_conv, dtype=np.float32))
                    mixed4_conv=[]
                    
                    #Create conv6
                    hdf5_f.create_dataset("conv6_"+str(batch_step), data = np.array(mixed6_conv, dtype=np.float32))
                    mixed6_conv=[]
                    
                    #Create conv8
                    hdf5_f.create_dataset("conv8_"+str(batch_step), data = np.array(mixed8_conv, dtype=np.float32))
                    mixed8_conv=[]

                    #Create conv9
                    hdf5_f.create_dataset("conv9_"+str(batch_step), data = np.array(mixed9_conv, dtype=np.float32))
                    mixed9_conv=[]
                    
                    #Create conv10
                    hdf5_f.create_dataset("conv10_"+str(batch_step), data = np.array(mixed10_conv, dtype=np.float32))
                    mixed10_conv=[]
                    
                    #Create conv10
                    hdf5_f.create_dataset("predictions_"+str(batch_step), data = np.array(predictions, dtype=np.float32))
                    predictions=[]
                    
                    batch_step += 1 

                    
                image_list += [img.split('/')[-1]]

                image_data = tf.gfile.FastGFile(img,'rb').read()

                mixed2_tensor_value = sess.run(mixed2_tensor, {'DecodeJpeg/contents:0': image_data})
                mixed2_conv += [mixed2_tensor_value[0,:,:,:]]

                mixed4_tensor_value = sess.run(mixed4_tensor, {'mixed_2/join:0': mixed2_tensor_value})
                mixed4_conv += [mixed4_tensor_value[0,:,:,:]]

                mixed6_tensor_value = sess.run(mixed6_tensor, {'mixed_4/join:0': mixed4_tensor_value})
                mixed6_conv += [mixed6_tensor_value[0,:,:,:]]

                mixed8_tensor_value = sess.run(mixed8_tensor, {'mixed_6/join:0': mixed6_tensor_value})
                mixed8_conv += [mixed8_tensor_value[0,:,:,:]]
                
                mixed9_tensor_value = sess.run(mixed9_tensor, {'mixed_8/join:0': mixed8_tensor_value})
                mixed9_conv += [mixed9_tensor_value[0,:,:,:]]
                
                mixed10_tensor_value = sess.run(mixed10_tensor, {'mixed_9/join:0': mixed9_tensor_value})
                mixed10_conv += [mixed10_tensor_value[0,:,:,:]]

                predictions += [sess.run(softmax_tensor,{'mixed_10/join:0': mixed10_tensor_value})[0,:]]

                
                if i % 500 == 0:
                    print('Processed', i, 'images from ',len(batch_img))
                    
            hdf5_f.create_dataset("image_list", data = image_list)
            hdf5_f.create_dataset("num_batches", data = batch_step)


    

In [16]:

    

#Save features as np arrays
def create_inception_features_hdf5(dir_images, dataset_type='trn', filter_image_type='*.jpg'):
    # List of images
    batch_img = []
    for root, dirnames, filenames in os.walk(dir_images):
        for filename in fnmatch.filter(filenames, filter_image_type):
            batch_img.append(os.path.join(root, filename))
    batch_img = sorted(batch_img)
            
    # Evaluate inception
    evaluate_inception_hdf5(batch_img, path + "data/feat_inception/"+dataset_type+".hdf5")
    
    print('Done creating numpy features for', dir_images)

# Train
create_inception_features_hdf5(path + 'data/imgs_trnval/train/',  dataset_type='trn')
create_inception_features_hdf5(path + 'data/imgs_trnval/val/',  dataset_type='tst')


    

    




('Processed', 0, 'images from ', 19654)
('Processed', 500, 'images from ', 19654)
('Processed', 1000, 'images from ', 19654)
('Processed', 1500, 'images from ', 19654)
('Processed', 2000, 'images from ', 19654)
('Processed', 2500, 'images from ', 19654)
('Processed', 3000, 'images from ', 19654)
('Processed', 3500, 'images from ', 19654)
('Processed', 4000, 'images from ', 19654)
('Processed', 4500, 'images from ', 19654)
('Processed', 5000, 'images from ', 19654)
('Processed', 5500, 'images from ', 19654)
('Processed', 6000, 'images from ', 19654)
('Processed', 6500, 'images from ', 19654)
('Processed', 7000, 'images from ', 19654)
('Processed', 7500, 'images from ', 19654)
('Processed', 8000, 'images from ', 19654)
('Processed', 8500, 'images from ', 19654)
('Processed', 9000, 'images from ', 19654)
('Processed', 9500, 'images from ', 19654)
('Processed', 10000, 'images from ', 19654)
('Processed', 10500, 'images from ', 19654)
('Processed', 11000, 'images from ', 19654)
('Processed', 11500, 'images from ', 19654)
('Processed', 12000, 'images from ', 19654)
('Processed', 12500, 'images from ', 19654)
('Processed', 13000, 'images from ', 19654)
('Processed', 13500, 'images from ', 19654)
('Processed', 14000, 'images from ', 19654)
('Processed', 14500, 'images from ', 19654)
('Processed', 15000, 'images from ', 19654)
('Processed', 15500, 'images from ', 19654)
('Processed', 16000, 'images from ', 19654)
('Processed', 16500, 'images from ', 19654)
('Processed', 17000, 'images from ', 19654)
('Processed', 17500, 'images from ', 19654)
('Processed', 18000, 'images from ', 19654)
('Processed', 18500, 'images from ', 19654)
('Processed', 19000, 'images from ', 19654)
('Processed', 19500, 'images from ', 19654)
('Done creating numpy features for', '/home/jorge/data/kaggle/statefarm/data/imgs_trnval/train/')
('Processed', 0, 'images from ', 2770)
('Processed', 500, 'images from ', 2770)
('Processed', 1000, 'images from ', 2770)
('Processed', 1500, 'images from ', 2770)
('Processed', 2000, 'images from ', 2770)
('Processed', 2500, 'images from ', 2770)
('Done creating numpy features for', '/home/jorge/data/kaggle/statefarm/data/imgs_trnval/val/')

In [35]:

    

#Check hd5 file
with h5py.File(path + "data/feat_inception/trn.hdf5", "r") as f:
    print(f.keys())
    num_batches = np.copy(f['num_batches'])
    
    print(num_batches)
    print(f['image_list'][:10])
    #X = hdf5_f["X"]


    

    




[u'conv2_0', u'conv2_1', u'conv2_2', u'conv2_3', u'conv2_4', u'conv2_5', u'conv2_6', u'conv2_7', u'image_list', u'num_batches']
8
['img_100026.jpg' 'img_10003.jpg' 'img_100050.jpg' 'img_100074.jpg'
 'img_100145.jpg' 'img_100191.jpg' 'img_100257.jpg' 'img_100312.jpg'
 'img_100337.jpg' 'img_100456.jpg']

In [16]:

    

import matplotlib.pyplot as plt
%matplotlib inline
plt.rcParams['figure.figsize'] = (10, 10)        # size of images

fig = plt.figure()
for i in xrange(25):
    a = fig.add_subplot(5,5,i)
    #a.set_title('Target: ' + str(np.argmax(mnist.train.labels[i])))
    fig.tight_layout()
    plt.imshow(conv2[0,:,:,i+50])


    

In [ ]:

    

#Save features as np arrays
def create_inception_features_array(dir_images, dataset_type='trn', filter_image_type='*.jpg'):
    # List of images
    batch_img = []
    for root, dirnames, filenames in os.walk(dir_images):
        for filename in fnmatch.filter(filenames, filter_image_type):
            batch_img.append(os.path.join(root, filename))
            
    # Evaluate inception
    image_list, conv2, conv8, layer8, conv9, layer9, conv10, layer10, predictions = evaluate_inception(batch_img)

    # Order by image name
    pd_list =  pd.DataFrame(image_list, columns=['img'])
    order = pd_list.sort('img').index
    
    np.save(path + "data/feat_inception/conv2."  + dataset_type, conv2[order])
    np.save(path + "data/feat_inception/conv8."  + dataset_type, conv8[order])
    np.save(path + "data/feat_inception/conv9."  + dataset_type, conv9[order])
    np.save(path + "data/feat_inception/conv10." + dataset_type, conv10[order])
    
    print('Done creating numpy features for', dir_images)


# Train
create_inception_features_array(path + 'data/imgs_trnval/train/',  dataset_type='trn')

# Test
create_inception_features_array(path + 'data/imgs_trnval/val/',  dataset_type='tst')

# score
create_inception_features_array(path + 'sources/imgs/test/',  dataset_type='sco')


    

    




('Processed', 0, 'images from ', 19654)
('Processed', 500, 'images from ', 19654)
('Processed', 1000, 'images from ', 19654)
('Processed', 1500, 'images from ', 19654)
('Processed', 2000, 'images from ', 19654)
('Processed', 2500, 'images from ', 19654)
('Processed', 3000, 'images from ', 19654)
('Processed', 3500, 'images from ', 19654)
('Processed', 4000, 'images from ', 19654)
('Processed', 4500, 'images from ', 19654)
('Processed', 5000, 'images from ', 19654)
('Processed', 5500, 'images from ', 19654)
('Processed', 6000, 'images from ', 19654)
('Processed', 6500, 'images from ', 19654)
('Processed', 7000, 'images from ', 19654)
('Processed', 7500, 'images from ', 19654)
('Processed', 8000, 'images from ', 19654)
('Processed', 8500, 'images from ', 19654)
('Processed', 9000, 'images from ', 19654)
('Processed', 9500, 'images from ', 19654)
('Processed', 10000, 'images from ', 19654)
('Processed', 10500, 'images from ', 19654)
('Processed', 11000, 'images from ', 19654)
('Processed', 11500, 'images from ', 19654)
('Processed', 12000, 'images from ', 19654)
('Processed', 12500, 'images from ', 19654)
('Processed', 13000, 'images from ', 19654)
('Processed', 13500, 'images from ', 19654)
('Processed', 14000, 'images from ', 19654)
('Processed', 14500, 'images from ', 19654)
('Processed', 15000, 'images from ', 19654)
('Processed', 15500, 'images from ', 19654)
('Processed', 16000, 'images from ', 19654)
('Processed', 16500, 'images from ', 19654)
('Processed', 17000, 'images from ', 19654)
('Processed', 17500, 'images from ', 19654)
('Processed', 18000, 'images from ', 19654)
('Processed', 18500, 'images from ', 19654)
('Processed', 19000, 'images from ', 19654)
('Processed', 19500, 'images from ', 19654)
('Done creating numpy features for', '/home/jorge/data/kaggle/statefarm/data/imgs_trnval/train/')
('Processed', 0, 'images from ', 2770)
('Processed', 500, 'images from ', 2770)
('Processed', 1000, 'images from ', 2770)
('Processed', 1500, 'images from ', 2770)
('Processed', 2000, 'images from ', 2770)
('Processed', 2500, 'images from ', 2770)
('Done creating numpy features for', '/home/jorge/data/kaggle/statefarm/data/imgs_trnval/val/')
('Processed', 0, 'images from ', 79726)
('Processed', 500, 'images from ', 79726)
('Processed', 1000, 'images from ', 79726)
('Processed', 1500, 'images from ', 79726)
('Processed', 2000, 'images from ', 79726)
('Processed', 2500, 'images from ', 79726)
('Processed', 3000, 'images from ', 79726)
('Processed', 3500, 'images from ', 79726)
('Processed', 4000, 'images from ', 79726)
('Processed', 4500, 'images from ', 79726)
('Processed', 5000, 'images from ', 79726)
('Processed', 5500, 'images from ', 79726)
('Processed', 6000, 'images from ', 79726)
('Processed', 6500, 'images from ', 79726)
('Processed', 7000, 'images from ', 79726)
('Processed', 7500, 'images from ', 79726)
('Processed', 8000, 'images from ', 79726)
('Processed', 8500, 'images from ', 79726)
('Processed', 9000, 'images from ', 79726)
('Processed', 9500, 'images from ', 79726)
('Processed', 10000, 'images from ', 79726)
('Processed', 10500, 'images from ', 79726)
('Processed', 11000, 'images from ', 79726)
('Processed', 11500, 'images from ', 79726)
('Processed', 12000, 'images from ', 79726)
('Processed', 12500, 'images from ', 79726)
('Processed', 13000, 'images from ', 79726)
('Processed', 13500, 'images from ', 79726)
('Processed', 14000, 'images from ', 79726)
('Processed', 14500, 'images from ', 79726)
('Processed', 15000, 'images from ', 79726)
('Processed', 15500, 'images from ', 79726)
('Processed', 16000, 'images from ', 79726)
('Processed', 16500, 'images from ', 79726)
('Processed', 17000, 'images from ', 79726)
('Processed', 17500, 'images from ', 79726)
('Processed', 18000, 'images from ', 79726)
('Processed', 18500, 'images from ', 79726)
('Processed', 19000, 'images from ', 79726)
('Processed', 19500, 'images from ', 79726)
('Processed', 20000, 'images from ', 79726)
('Processed', 20500, 'images from ', 79726)
('Processed', 21000, 'images from ', 79726)
('Processed', 21500, 'images from ', 79726)
('Processed', 22000, 'images from ', 79726)
('Processed', 22500, 'images from ', 79726)
('Processed', 23000, 'images from ', 79726)
('Processed', 23500, 'images from ', 79726)
('Processed', 24000, 'images from ', 79726)
('Processed', 24500, 'images from ', 79726)
('Processed', 25000, 'images from ', 79726)
('Processed', 25500, 'images from ', 79726)
('Processed', 26000, 'images from ', 79726)
('Processed', 26500, 'images from ', 79726)
('Processed', 27000, 'images from ', 79726)
('Processed', 27500, 'images from ', 79726)
('Processed', 28000, 'images from ', 79726)
('Processed', 28500, 'images from ', 79726)
('Processed', 29000, 'images from ', 79726)
('Processed', 29500, 'images from ', 79726)
('Processed', 30000, 'images from ', 79726)
('Processed', 30500, 'images from ', 79726)
('Processed', 31000, 'images from ', 79726)
('Processed', 31500, 'images from ', 79726)
('Processed', 32000, 'images from ', 79726)
('Processed', 32500, 'images from ', 79726)
('Processed', 33000, 'images from ', 79726)
('Processed', 33500, 'images from ', 79726)
('Processed', 34000, 'images from ', 79726)
('Processed', 34500, 'images from ', 79726)
('Processed', 35000, 'images from ', 79726)
('Processed', 35500, 'images from ', 79726)
('Processed', 36000, 'images from ', 79726)
('Processed', 36500, 'images from ', 79726)
('Processed', 37000, 'images from ', 79726)
('Processed', 37500, 'images from ', 79726)
('Processed', 38000, 'images from ', 79726)
('Processed', 38500, 'images from ', 79726)
('Processed', 39000, 'images from ', 79726)
('Processed', 39500, 'images from ', 79726)
('Processed', 40000, 'images from ', 79726)
('Processed', 40500, 'images from ', 79726)
('Processed', 41000, 'images from ', 79726)
('Processed', 41500, 'images from ', 79726)
('Processed', 42000, 'images from ', 79726)
('Processed', 42500, 'images from ', 79726)
('Processed', 43000, 'images from ', 79726)
('Processed', 43500, 'images from ', 79726)
('Processed', 44000, 'images from ', 79726)
('Processed', 44500, 'images from ', 79726)
('Processed', 45000, 'images from ', 79726)
('Processed', 45500, 'images from ', 79726)

In [8]:

    

#Save features as csv
def create_inception_features(dir_images, dataset_type='trn', filter_image_type='*.jpg'):
    # List of images
    batch_img = []
    for root, dirnames, filenames in os.walk(dir_images):
        for filename in fnmatch.filter(filenames, filter_image_type):
            batch_img.append(os.path.join(root, filename))
            
    # Evaluate inception
    image_list, layer8, layer9, layer10, predictions = evaluate_inception(batch_img)

    #Generate csv files for features
    layer8_cols =      ['incepcion_ly8_'  + str(i) for i in range(layer8.shape[1])]
    df = pd.DataFrame(layer8, columns=layer8_cols, index=image_list)
    df.sort_index(inplace=True)
    df.to_csv(path + "data/feat_inception/layer8.csv." + dataset_type)
    
    layer9_cols =      ['incepcion_ly9_'  + str(i) for i in range(layer9.shape[1])]
    df = pd.DataFrame(layer9, columns=layer9_cols, index=image_list)
    df.sort_index(inplace=True)
    df.to_csv(path + "data/feat_inception/layer9.csv." + dataset_type)
    
    layer10_cols =     ['incepcion_ly10_' + str(i) for i in range(layer10.shape[1])]
    df = pd.DataFrame(layer10, columns=layer10_cols, index=image_list)
    df.sort_index(inplace=True)
    df.to_csv(path + "data/feat_inception/layer10.csv." + dataset_type)
    
    predictions_cols = ['incepcion_pred_' + str(i) for i in range(predictions.shape[1])]
    df = pd.DataFrame(predictions, columns=predictions_cols, index=image_list)
    df.sort_index(inplace=True)
    df.to_csv(path + "data/feat_inception/predictions.csv." + dataset_type)

    print('Done creating features for', dir_images)


# Train
#create_inception_features(path + '/data/imgs_trnval/train/',  dataset_type='trn')

# Test
create_inception_features(path + '/data/imgs_trnval/val/',  dataset_type='tst')

# score
#create_inception_features(path + '/sources/imgs/score',  dataset_type='sco')


    

    




('Processed', 0, 'images from ', 2770)
('Processed', 500, 'images from ', 2770)
('Processed', 1000, 'images from ', 2770)
('Processed', 1500, 'images from ', 2770)
('Processed', 2000, 'images from ', 2770)
('Processed', 2500, 'images from ', 2770)
('Done creating features for', '/home/jorge/data/kaggle/statefarm//data/imgs_trnval/val/')

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [5]:

    

#List all layers of a net
with tf.Session() as sess:
    for op in sess.graph.get_operations():
        print(op.name)
        print(op.values())


    

    




DecodeJpeg/contents
(<tf.Tensor 'DecodeJpeg/contents:0' shape=() dtype=string>,)
DecodeJpeg
(<tf.Tensor 'DecodeJpeg:0' shape=(?, ?, 3) dtype=uint8>,)
Cast
(<tf.Tensor 'Cast:0' shape=(?, ?, 3) dtype=float32>,)
ExpandDims/dim
(<tf.Tensor 'ExpandDims/dim:0' shape=(1,) dtype=int32>,)
ExpandDims
(<tf.Tensor 'ExpandDims:0' shape=(1, ?, ?, 3) dtype=float32>,)
ResizeBilinear/size
(<tf.Tensor 'ResizeBilinear/size:0' shape=(2,) dtype=int32>,)
ResizeBilinear
(<tf.Tensor 'ResizeBilinear:0' shape=(1, 299, 299, 3) dtype=float32>,)
Sub/y
(<tf.Tensor 'Sub/y:0' shape=() dtype=float32>,)
Sub
(<tf.Tensor 'Sub:0' shape=(1, 299, 299, 3) dtype=float32>,)
Mul/y
(<tf.Tensor 'Mul/y:0' shape=() dtype=float32>,)
Mul
(<tf.Tensor 'Mul:0' shape=(1, 299, 299, 3) dtype=float32>,)
conv/conv2d_params
(<tf.Tensor 'conv/conv2d_params:0' shape=(3, 3, 3, 32) dtype=float32>,)
conv/Conv2D
(<tf.Tensor 'conv/Conv2D:0' shape=(1, 149, 149, 32) dtype=float32>,)
conv/batchnorm/beta
(<tf.Tensor 'conv/batchnorm/beta:0' shape=(32,) dtype=float32>,)
conv/batchnorm/gamma
(<tf.Tensor 'conv/batchnorm/gamma:0' shape=(32,) dtype=float32>,)
conv/batchnorm/moving_mean
(<tf.Tensor 'conv/batchnorm/moving_mean:0' shape=(32,) dtype=float32>,)
conv/batchnorm/moving_variance
(<tf.Tensor 'conv/batchnorm/moving_variance:0' shape=(32,) dtype=float32>,)
conv/batchnorm
(<tf.Tensor 'conv/batchnorm:0' shape=(1, 149, 149, 32) dtype=float32>,)
conv/CheckNumerics
(<tf.Tensor 'conv/CheckNumerics:0' shape=(1, 149, 149, 32) dtype=float32>,)
conv/control_dependency
(<tf.Tensor 'conv/control_dependency:0' shape=(1, 149, 149, 32) dtype=float32>,)
conv
(<tf.Tensor 'conv:0' shape=(1, 149, 149, 32) dtype=float32>,)
conv_1/conv2d_params
(<tf.Tensor 'conv_1/conv2d_params:0' shape=(3, 3, 32, 32) dtype=float32>,)
conv_1/Conv2D
(<tf.Tensor 'conv_1/Conv2D:0' shape=(1, 147, 147, 32) dtype=float32>,)
conv_1/batchnorm/beta
(<tf.Tensor 'conv_1/batchnorm/beta:0' shape=(32,) dtype=float32>,)
conv_1/batchnorm/gamma
(<tf.Tensor 'conv_1/batchnorm/gamma:0' shape=(32,) dtype=float32>,)
conv_1/batchnorm/moving_mean
(<tf.Tensor 'conv_1/batchnorm/moving_mean:0' shape=(32,) dtype=float32>,)
conv_1/batchnorm/moving_variance
(<tf.Tensor 'conv_1/batchnorm/moving_variance:0' shape=(32,) dtype=float32>,)
conv_1/batchnorm
(<tf.Tensor 'conv_1/batchnorm:0' shape=(1, 147, 147, 32) dtype=float32>,)
conv_1/CheckNumerics
(<tf.Tensor 'conv_1/CheckNumerics:0' shape=(1, 147, 147, 32) dtype=float32>,)
conv_1/control_dependency
(<tf.Tensor 'conv_1/control_dependency:0' shape=(1, 147, 147, 32) dtype=float32>,)
conv_1
(<tf.Tensor 'conv_1:0' shape=(1, 147, 147, 32) dtype=float32>,)
conv_2/conv2d_params
(<tf.Tensor 'conv_2/conv2d_params:0' shape=(3, 3, 32, 64) dtype=float32>,)
conv_2/Conv2D
(<tf.Tensor 'conv_2/Conv2D:0' shape=(1, 147, 147, 64) dtype=float32>,)
conv_2/batchnorm/beta
(<tf.Tensor 'conv_2/batchnorm/beta:0' shape=(64,) dtype=float32>,)
conv_2/batchnorm/gamma
(<tf.Tensor 'conv_2/batchnorm/gamma:0' shape=(64,) dtype=float32>,)
conv_2/batchnorm/moving_mean
(<tf.Tensor 'conv_2/batchnorm/moving_mean:0' shape=(64,) dtype=float32>,)
conv_2/batchnorm/moving_variance
(<tf.Tensor 'conv_2/batchnorm/moving_variance:0' shape=(64,) dtype=float32>,)
conv_2/batchnorm
(<tf.Tensor 'conv_2/batchnorm:0' shape=(1, 147, 147, 64) dtype=float32>,)
conv_2/CheckNumerics
(<tf.Tensor 'conv_2/CheckNumerics:0' shape=(1, 147, 147, 64) dtype=float32>,)
conv_2/control_dependency
(<tf.Tensor 'conv_2/control_dependency:0' shape=(1, 147, 147, 64) dtype=float32>,)
conv_2
(<tf.Tensor 'conv_2:0' shape=(1, 147, 147, 64) dtype=float32>,)
pool/CheckNumerics
(<tf.Tensor 'pool/CheckNumerics:0' shape=(1, 147, 147, 64) dtype=float32>,)
pool/control_dependency
(<tf.Tensor 'pool/control_dependency:0' shape=(1, 147, 147, 64) dtype=float32>,)
pool
(<tf.Tensor 'pool:0' shape=(1, 73, 73, 64) dtype=float32>,)
conv_3/conv2d_params
(<tf.Tensor 'conv_3/conv2d_params:0' shape=(1, 1, 64, 80) dtype=float32>,)
conv_3/Conv2D
(<tf.Tensor 'conv_3/Conv2D:0' shape=(1, 73, 73, 80) dtype=float32>,)
conv_3/batchnorm/beta
(<tf.Tensor 'conv_3/batchnorm/beta:0' shape=(80,) dtype=float32>,)
conv_3/batchnorm/gamma
(<tf.Tensor 'conv_3/batchnorm/gamma:0' shape=(80,) dtype=float32>,)
conv_3/batchnorm/moving_mean
(<tf.Tensor 'conv_3/batchnorm/moving_mean:0' shape=(80,) dtype=float32>,)
conv_3/batchnorm/moving_variance
(<tf.Tensor 'conv_3/batchnorm/moving_variance:0' shape=(80,) dtype=float32>,)
conv_3/batchnorm
(<tf.Tensor 'conv_3/batchnorm:0' shape=(1, 73, 73, 80) dtype=float32>,)
conv_3/CheckNumerics
(<tf.Tensor 'conv_3/CheckNumerics:0' shape=(1, 73, 73, 80) dtype=float32>,)
conv_3/control_dependency
(<tf.Tensor 'conv_3/control_dependency:0' shape=(1, 73, 73, 80) dtype=float32>,)
conv_3
(<tf.Tensor 'conv_3:0' shape=(1, 73, 73, 80) dtype=float32>,)
conv_4/conv2d_params
(<tf.Tensor 'conv_4/conv2d_params:0' shape=(3, 3, 80, 192) dtype=float32>,)
conv_4/Conv2D
(<tf.Tensor 'conv_4/Conv2D:0' shape=(1, 71, 71, 192) dtype=float32>,)
conv_4/batchnorm/beta
(<tf.Tensor 'conv_4/batchnorm/beta:0' shape=(192,) dtype=float32>,)
conv_4/batchnorm/gamma
(<tf.Tensor 'conv_4/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
conv_4/batchnorm/moving_mean
(<tf.Tensor 'conv_4/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
conv_4/batchnorm/moving_variance
(<tf.Tensor 'conv_4/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
conv_4/batchnorm
(<tf.Tensor 'conv_4/batchnorm:0' shape=(1, 71, 71, 192) dtype=float32>,)
conv_4/CheckNumerics
(<tf.Tensor 'conv_4/CheckNumerics:0' shape=(1, 71, 71, 192) dtype=float32>,)
conv_4/control_dependency
(<tf.Tensor 'conv_4/control_dependency:0' shape=(1, 71, 71, 192) dtype=float32>,)
conv_4
(<tf.Tensor 'conv_4:0' shape=(1, 71, 71, 192) dtype=float32>,)
pool_1/CheckNumerics
(<tf.Tensor 'pool_1/CheckNumerics:0' shape=(1, 71, 71, 192) dtype=float32>,)
pool_1/control_dependency
(<tf.Tensor 'pool_1/control_dependency:0' shape=(1, 71, 71, 192) dtype=float32>,)
pool_1
(<tf.Tensor 'pool_1:0' shape=(1, 35, 35, 192) dtype=float32>,)
mixed/conv/conv2d_params
(<tf.Tensor 'mixed/conv/conv2d_params:0' shape=(1, 1, 192, 64) dtype=float32>,)
mixed/conv/Conv2D
(<tf.Tensor 'mixed/conv/Conv2D:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/conv/batchnorm/beta
(<tf.Tensor 'mixed/conv/batchnorm/beta:0' shape=(64,) dtype=float32>,)
mixed/conv/batchnorm/gamma
(<tf.Tensor 'mixed/conv/batchnorm/gamma:0' shape=(64,) dtype=float32>,)
mixed/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed/conv/batchnorm/moving_mean:0' shape=(64,) dtype=float32>,)
mixed/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed/conv/batchnorm/moving_variance:0' shape=(64,) dtype=float32>,)
mixed/conv/batchnorm
(<tf.Tensor 'mixed/conv/batchnorm:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/conv/CheckNumerics
(<tf.Tensor 'mixed/conv/CheckNumerics:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/conv/control_dependency
(<tf.Tensor 'mixed/conv/control_dependency:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/conv
(<tf.Tensor 'mixed/conv:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/tower/conv/conv2d_params
(<tf.Tensor 'mixed/tower/conv/conv2d_params:0' shape=(1, 1, 192, 48) dtype=float32>,)
mixed/tower/conv/Conv2D
(<tf.Tensor 'mixed/tower/conv/Conv2D:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed/tower/conv/batchnorm/beta
(<tf.Tensor 'mixed/tower/conv/batchnorm/beta:0' shape=(48,) dtype=float32>,)
mixed/tower/conv/batchnorm/gamma
(<tf.Tensor 'mixed/tower/conv/batchnorm/gamma:0' shape=(48,) dtype=float32>,)
mixed/tower/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed/tower/conv/batchnorm/moving_mean:0' shape=(48,) dtype=float32>,)
mixed/tower/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed/tower/conv/batchnorm/moving_variance:0' shape=(48,) dtype=float32>,)
mixed/tower/conv/batchnorm
(<tf.Tensor 'mixed/tower/conv/batchnorm:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed/tower/conv/CheckNumerics
(<tf.Tensor 'mixed/tower/conv/CheckNumerics:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed/tower/conv/control_dependency
(<tf.Tensor 'mixed/tower/conv/control_dependency:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed/tower/conv
(<tf.Tensor 'mixed/tower/conv:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed/tower/conv_1/conv2d_params
(<tf.Tensor 'mixed/tower/conv_1/conv2d_params:0' shape=(5, 5, 48, 64) dtype=float32>,)
mixed/tower/conv_1/Conv2D
(<tf.Tensor 'mixed/tower/conv_1/Conv2D:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/tower/conv_1/batchnorm/beta
(<tf.Tensor 'mixed/tower/conv_1/batchnorm/beta:0' shape=(64,) dtype=float32>,)
mixed/tower/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed/tower/conv_1/batchnorm/gamma:0' shape=(64,) dtype=float32>,)
mixed/tower/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed/tower/conv_1/batchnorm/moving_mean:0' shape=(64,) dtype=float32>,)
mixed/tower/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed/tower/conv_1/batchnorm/moving_variance:0' shape=(64,) dtype=float32>,)
mixed/tower/conv_1/batchnorm
(<tf.Tensor 'mixed/tower/conv_1/batchnorm:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/tower/conv_1/CheckNumerics
(<tf.Tensor 'mixed/tower/conv_1/CheckNumerics:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/tower/conv_1/control_dependency
(<tf.Tensor 'mixed/tower/conv_1/control_dependency:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/tower/conv_1
(<tf.Tensor 'mixed/tower/conv_1:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/tower_1/conv/conv2d_params
(<tf.Tensor 'mixed/tower_1/conv/conv2d_params:0' shape=(1, 1, 192, 64) dtype=float32>,)
mixed/tower_1/conv/Conv2D
(<tf.Tensor 'mixed/tower_1/conv/Conv2D:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/tower_1/conv/batchnorm/beta
(<tf.Tensor 'mixed/tower_1/conv/batchnorm/beta:0' shape=(64,) dtype=float32>,)
mixed/tower_1/conv/batchnorm/gamma
(<tf.Tensor 'mixed/tower_1/conv/batchnorm/gamma:0' shape=(64,) dtype=float32>,)
mixed/tower_1/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed/tower_1/conv/batchnorm/moving_mean:0' shape=(64,) dtype=float32>,)
mixed/tower_1/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed/tower_1/conv/batchnorm/moving_variance:0' shape=(64,) dtype=float32>,)
mixed/tower_1/conv/batchnorm
(<tf.Tensor 'mixed/tower_1/conv/batchnorm:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/tower_1/conv/CheckNumerics
(<tf.Tensor 'mixed/tower_1/conv/CheckNumerics:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/tower_1/conv/control_dependency
(<tf.Tensor 'mixed/tower_1/conv/control_dependency:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/tower_1/conv
(<tf.Tensor 'mixed/tower_1/conv:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed/tower_1/conv_1/conv2d_params
(<tf.Tensor 'mixed/tower_1/conv_1/conv2d_params:0' shape=(3, 3, 64, 96) dtype=float32>,)
mixed/tower_1/conv_1/Conv2D
(<tf.Tensor 'mixed/tower_1/conv_1/Conv2D:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed/tower_1/conv_1/batchnorm/beta
(<tf.Tensor 'mixed/tower_1/conv_1/batchnorm/beta:0' shape=(96,) dtype=float32>,)
mixed/tower_1/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed/tower_1/conv_1/batchnorm/gamma:0' shape=(96,) dtype=float32>,)
mixed/tower_1/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed/tower_1/conv_1/batchnorm/moving_mean:0' shape=(96,) dtype=float32>,)
mixed/tower_1/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed/tower_1/conv_1/batchnorm/moving_variance:0' shape=(96,) dtype=float32>,)
mixed/tower_1/conv_1/batchnorm
(<tf.Tensor 'mixed/tower_1/conv_1/batchnorm:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed/tower_1/conv_1/CheckNumerics
(<tf.Tensor 'mixed/tower_1/conv_1/CheckNumerics:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed/tower_1/conv_1/control_dependency
(<tf.Tensor 'mixed/tower_1/conv_1/control_dependency:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed/tower_1/conv_1
(<tf.Tensor 'mixed/tower_1/conv_1:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed/tower_1/conv_2/conv2d_params
(<tf.Tensor 'mixed/tower_1/conv_2/conv2d_params:0' shape=(3, 3, 96, 96) dtype=float32>,)
mixed/tower_1/conv_2/Conv2D
(<tf.Tensor 'mixed/tower_1/conv_2/Conv2D:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed/tower_1/conv_2/batchnorm/beta
(<tf.Tensor 'mixed/tower_1/conv_2/batchnorm/beta:0' shape=(96,) dtype=float32>,)
mixed/tower_1/conv_2/batchnorm/gamma
(<tf.Tensor 'mixed/tower_1/conv_2/batchnorm/gamma:0' shape=(96,) dtype=float32>,)
mixed/tower_1/conv_2/batchnorm/moving_mean
(<tf.Tensor 'mixed/tower_1/conv_2/batchnorm/moving_mean:0' shape=(96,) dtype=float32>,)
mixed/tower_1/conv_2/batchnorm/moving_variance
(<tf.Tensor 'mixed/tower_1/conv_2/batchnorm/moving_variance:0' shape=(96,) dtype=float32>,)
mixed/tower_1/conv_2/batchnorm
(<tf.Tensor 'mixed/tower_1/conv_2/batchnorm:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed/tower_1/conv_2/CheckNumerics
(<tf.Tensor 'mixed/tower_1/conv_2/CheckNumerics:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed/tower_1/conv_2/control_dependency
(<tf.Tensor 'mixed/tower_1/conv_2/control_dependency:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed/tower_1/conv_2
(<tf.Tensor 'mixed/tower_1/conv_2:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed/tower_2/pool
(<tf.Tensor 'mixed/tower_2/pool:0' shape=(1, 35, 35, 192) dtype=float32>,)
mixed/tower_2/conv/conv2d_params
(<tf.Tensor 'mixed/tower_2/conv/conv2d_params:0' shape=(1, 1, 192, 32) dtype=float32>,)
mixed/tower_2/conv/Conv2D
(<tf.Tensor 'mixed/tower_2/conv/Conv2D:0' shape=(1, 35, 35, 32) dtype=float32>,)
mixed/tower_2/conv/batchnorm/beta
(<tf.Tensor 'mixed/tower_2/conv/batchnorm/beta:0' shape=(32,) dtype=float32>,)
mixed/tower_2/conv/batchnorm/gamma
(<tf.Tensor 'mixed/tower_2/conv/batchnorm/gamma:0' shape=(32,) dtype=float32>,)
mixed/tower_2/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed/tower_2/conv/batchnorm/moving_mean:0' shape=(32,) dtype=float32>,)
mixed/tower_2/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed/tower_2/conv/batchnorm/moving_variance:0' shape=(32,) dtype=float32>,)
mixed/tower_2/conv/batchnorm
(<tf.Tensor 'mixed/tower_2/conv/batchnorm:0' shape=(1, 35, 35, 32) dtype=float32>,)
mixed/tower_2/conv/CheckNumerics
(<tf.Tensor 'mixed/tower_2/conv/CheckNumerics:0' shape=(1, 35, 35, 32) dtype=float32>,)
mixed/tower_2/conv/control_dependency
(<tf.Tensor 'mixed/tower_2/conv/control_dependency:0' shape=(1, 35, 35, 32) dtype=float32>,)
mixed/tower_2/conv
(<tf.Tensor 'mixed/tower_2/conv:0' shape=(1, 35, 35, 32) dtype=float32>,)
mixed/join/concat_dim
(<tf.Tensor 'mixed/join/concat_dim:0' shape=() dtype=int32>,)
mixed/join
(<tf.Tensor 'mixed/join:0' shape=(1, 35, 35, 256) dtype=float32>,)
mixed_1/conv/conv2d_params
(<tf.Tensor 'mixed_1/conv/conv2d_params:0' shape=(1, 1, 256, 64) dtype=float32>,)
mixed_1/conv/Conv2D
(<tf.Tensor 'mixed_1/conv/Conv2D:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/conv/batchnorm/beta
(<tf.Tensor 'mixed_1/conv/batchnorm/beta:0' shape=(64,) dtype=float32>,)
mixed_1/conv/batchnorm/gamma
(<tf.Tensor 'mixed_1/conv/batchnorm/gamma:0' shape=(64,) dtype=float32>,)
mixed_1/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_1/conv/batchnorm/moving_mean:0' shape=(64,) dtype=float32>,)
mixed_1/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_1/conv/batchnorm/moving_variance:0' shape=(64,) dtype=float32>,)
mixed_1/conv/batchnorm
(<tf.Tensor 'mixed_1/conv/batchnorm:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/conv/CheckNumerics
(<tf.Tensor 'mixed_1/conv/CheckNumerics:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/conv/control_dependency
(<tf.Tensor 'mixed_1/conv/control_dependency:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/conv
(<tf.Tensor 'mixed_1/conv:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower/conv/conv2d_params
(<tf.Tensor 'mixed_1/tower/conv/conv2d_params:0' shape=(1, 1, 256, 48) dtype=float32>,)
mixed_1/tower/conv/Conv2D
(<tf.Tensor 'mixed_1/tower/conv/Conv2D:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed_1/tower/conv/batchnorm/beta
(<tf.Tensor 'mixed_1/tower/conv/batchnorm/beta:0' shape=(48,) dtype=float32>,)
mixed_1/tower/conv/batchnorm/gamma
(<tf.Tensor 'mixed_1/tower/conv/batchnorm/gamma:0' shape=(48,) dtype=float32>,)
mixed_1/tower/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_1/tower/conv/batchnorm/moving_mean:0' shape=(48,) dtype=float32>,)
mixed_1/tower/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_1/tower/conv/batchnorm/moving_variance:0' shape=(48,) dtype=float32>,)
mixed_1/tower/conv/batchnorm
(<tf.Tensor 'mixed_1/tower/conv/batchnorm:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed_1/tower/conv/CheckNumerics
(<tf.Tensor 'mixed_1/tower/conv/CheckNumerics:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed_1/tower/conv/control_dependency
(<tf.Tensor 'mixed_1/tower/conv/control_dependency:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed_1/tower/conv
(<tf.Tensor 'mixed_1/tower/conv:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed_1/tower/conv_1/conv2d_params
(<tf.Tensor 'mixed_1/tower/conv_1/conv2d_params:0' shape=(5, 5, 48, 64) dtype=float32>,)
mixed_1/tower/conv_1/Conv2D
(<tf.Tensor 'mixed_1/tower/conv_1/Conv2D:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_1/tower/conv_1/batchnorm/beta:0' shape=(64,) dtype=float32>,)
mixed_1/tower/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_1/tower/conv_1/batchnorm/gamma:0' shape=(64,) dtype=float32>,)
mixed_1/tower/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_1/tower/conv_1/batchnorm/moving_mean:0' shape=(64,) dtype=float32>,)
mixed_1/tower/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_1/tower/conv_1/batchnorm/moving_variance:0' shape=(64,) dtype=float32>,)
mixed_1/tower/conv_1/batchnorm
(<tf.Tensor 'mixed_1/tower/conv_1/batchnorm:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower/conv_1/CheckNumerics
(<tf.Tensor 'mixed_1/tower/conv_1/CheckNumerics:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower/conv_1/control_dependency
(<tf.Tensor 'mixed_1/tower/conv_1/control_dependency:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower/conv_1
(<tf.Tensor 'mixed_1/tower/conv_1:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower_1/conv/conv2d_params
(<tf.Tensor 'mixed_1/tower_1/conv/conv2d_params:0' shape=(1, 1, 256, 64) dtype=float32>,)
mixed_1/tower_1/conv/Conv2D
(<tf.Tensor 'mixed_1/tower_1/conv/Conv2D:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower_1/conv/batchnorm/beta
(<tf.Tensor 'mixed_1/tower_1/conv/batchnorm/beta:0' shape=(64,) dtype=float32>,)
mixed_1/tower_1/conv/batchnorm/gamma
(<tf.Tensor 'mixed_1/tower_1/conv/batchnorm/gamma:0' shape=(64,) dtype=float32>,)
mixed_1/tower_1/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_1/tower_1/conv/batchnorm/moving_mean:0' shape=(64,) dtype=float32>,)
mixed_1/tower_1/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_1/tower_1/conv/batchnorm/moving_variance:0' shape=(64,) dtype=float32>,)
mixed_1/tower_1/conv/batchnorm
(<tf.Tensor 'mixed_1/tower_1/conv/batchnorm:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower_1/conv/CheckNumerics
(<tf.Tensor 'mixed_1/tower_1/conv/CheckNumerics:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower_1/conv/control_dependency
(<tf.Tensor 'mixed_1/tower_1/conv/control_dependency:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower_1/conv
(<tf.Tensor 'mixed_1/tower_1/conv:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower_1/conv_1/conv2d_params
(<tf.Tensor 'mixed_1/tower_1/conv_1/conv2d_params:0' shape=(3, 3, 64, 96) dtype=float32>,)
mixed_1/tower_1/conv_1/Conv2D
(<tf.Tensor 'mixed_1/tower_1/conv_1/Conv2D:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_1/tower_1/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_1/tower_1/conv_1/batchnorm/beta:0' shape=(96,) dtype=float32>,)
mixed_1/tower_1/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_1/tower_1/conv_1/batchnorm/gamma:0' shape=(96,) dtype=float32>,)
mixed_1/tower_1/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_1/tower_1/conv_1/batchnorm/moving_mean:0' shape=(96,) dtype=float32>,)
mixed_1/tower_1/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_1/tower_1/conv_1/batchnorm/moving_variance:0' shape=(96,) dtype=float32>,)
mixed_1/tower_1/conv_1/batchnorm
(<tf.Tensor 'mixed_1/tower_1/conv_1/batchnorm:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_1/tower_1/conv_1/CheckNumerics
(<tf.Tensor 'mixed_1/tower_1/conv_1/CheckNumerics:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_1/tower_1/conv_1/control_dependency
(<tf.Tensor 'mixed_1/tower_1/conv_1/control_dependency:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_1/tower_1/conv_1
(<tf.Tensor 'mixed_1/tower_1/conv_1:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_1/tower_1/conv_2/conv2d_params
(<tf.Tensor 'mixed_1/tower_1/conv_2/conv2d_params:0' shape=(3, 3, 96, 96) dtype=float32>,)
mixed_1/tower_1/conv_2/Conv2D
(<tf.Tensor 'mixed_1/tower_1/conv_2/Conv2D:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_1/tower_1/conv_2/batchnorm/beta
(<tf.Tensor 'mixed_1/tower_1/conv_2/batchnorm/beta:0' shape=(96,) dtype=float32>,)
mixed_1/tower_1/conv_2/batchnorm/gamma
(<tf.Tensor 'mixed_1/tower_1/conv_2/batchnorm/gamma:0' shape=(96,) dtype=float32>,)
mixed_1/tower_1/conv_2/batchnorm/moving_mean
(<tf.Tensor 'mixed_1/tower_1/conv_2/batchnorm/moving_mean:0' shape=(96,) dtype=float32>,)
mixed_1/tower_1/conv_2/batchnorm/moving_variance
(<tf.Tensor 'mixed_1/tower_1/conv_2/batchnorm/moving_variance:0' shape=(96,) dtype=float32>,)
mixed_1/tower_1/conv_2/batchnorm
(<tf.Tensor 'mixed_1/tower_1/conv_2/batchnorm:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_1/tower_1/conv_2/CheckNumerics
(<tf.Tensor 'mixed_1/tower_1/conv_2/CheckNumerics:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_1/tower_1/conv_2/control_dependency
(<tf.Tensor 'mixed_1/tower_1/conv_2/control_dependency:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_1/tower_1/conv_2
(<tf.Tensor 'mixed_1/tower_1/conv_2:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_1/tower_2/pool
(<tf.Tensor 'mixed_1/tower_2/pool:0' shape=(1, 35, 35, 256) dtype=float32>,)
mixed_1/tower_2/conv/conv2d_params
(<tf.Tensor 'mixed_1/tower_2/conv/conv2d_params:0' shape=(1, 1, 256, 64) dtype=float32>,)
mixed_1/tower_2/conv/Conv2D
(<tf.Tensor 'mixed_1/tower_2/conv/Conv2D:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower_2/conv/batchnorm/beta
(<tf.Tensor 'mixed_1/tower_2/conv/batchnorm/beta:0' shape=(64,) dtype=float32>,)
mixed_1/tower_2/conv/batchnorm/gamma
(<tf.Tensor 'mixed_1/tower_2/conv/batchnorm/gamma:0' shape=(64,) dtype=float32>,)
mixed_1/tower_2/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_1/tower_2/conv/batchnorm/moving_mean:0' shape=(64,) dtype=float32>,)
mixed_1/tower_2/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_1/tower_2/conv/batchnorm/moving_variance:0' shape=(64,) dtype=float32>,)
mixed_1/tower_2/conv/batchnorm
(<tf.Tensor 'mixed_1/tower_2/conv/batchnorm:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower_2/conv/CheckNumerics
(<tf.Tensor 'mixed_1/tower_2/conv/CheckNumerics:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower_2/conv/control_dependency
(<tf.Tensor 'mixed_1/tower_2/conv/control_dependency:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/tower_2/conv
(<tf.Tensor 'mixed_1/tower_2/conv:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_1/join/concat_dim
(<tf.Tensor 'mixed_1/join/concat_dim:0' shape=() dtype=int32>,)
mixed_1/join
(<tf.Tensor 'mixed_1/join:0' shape=(1, 35, 35, 288) dtype=float32>,)
mixed_2/conv/conv2d_params
(<tf.Tensor 'mixed_2/conv/conv2d_params:0' shape=(1, 1, 288, 64) dtype=float32>,)
mixed_2/conv/Conv2D
(<tf.Tensor 'mixed_2/conv/Conv2D:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/conv/batchnorm/beta
(<tf.Tensor 'mixed_2/conv/batchnorm/beta:0' shape=(64,) dtype=float32>,)
mixed_2/conv/batchnorm/gamma
(<tf.Tensor 'mixed_2/conv/batchnorm/gamma:0' shape=(64,) dtype=float32>,)
mixed_2/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_2/conv/batchnorm/moving_mean:0' shape=(64,) dtype=float32>,)
mixed_2/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_2/conv/batchnorm/moving_variance:0' shape=(64,) dtype=float32>,)
mixed_2/conv/batchnorm
(<tf.Tensor 'mixed_2/conv/batchnorm:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/conv/CheckNumerics
(<tf.Tensor 'mixed_2/conv/CheckNumerics:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/conv/control_dependency
(<tf.Tensor 'mixed_2/conv/control_dependency:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/conv
(<tf.Tensor 'mixed_2/conv:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower/conv/conv2d_params
(<tf.Tensor 'mixed_2/tower/conv/conv2d_params:0' shape=(1, 1, 288, 48) dtype=float32>,)
mixed_2/tower/conv/Conv2D
(<tf.Tensor 'mixed_2/tower/conv/Conv2D:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed_2/tower/conv/batchnorm/beta
(<tf.Tensor 'mixed_2/tower/conv/batchnorm/beta:0' shape=(48,) dtype=float32>,)
mixed_2/tower/conv/batchnorm/gamma
(<tf.Tensor 'mixed_2/tower/conv/batchnorm/gamma:0' shape=(48,) dtype=float32>,)
mixed_2/tower/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_2/tower/conv/batchnorm/moving_mean:0' shape=(48,) dtype=float32>,)
mixed_2/tower/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_2/tower/conv/batchnorm/moving_variance:0' shape=(48,) dtype=float32>,)
mixed_2/tower/conv/batchnorm
(<tf.Tensor 'mixed_2/tower/conv/batchnorm:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed_2/tower/conv/CheckNumerics
(<tf.Tensor 'mixed_2/tower/conv/CheckNumerics:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed_2/tower/conv/control_dependency
(<tf.Tensor 'mixed_2/tower/conv/control_dependency:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed_2/tower/conv
(<tf.Tensor 'mixed_2/tower/conv:0' shape=(1, 35, 35, 48) dtype=float32>,)
mixed_2/tower/conv_1/conv2d_params
(<tf.Tensor 'mixed_2/tower/conv_1/conv2d_params:0' shape=(5, 5, 48, 64) dtype=float32>,)
mixed_2/tower/conv_1/Conv2D
(<tf.Tensor 'mixed_2/tower/conv_1/Conv2D:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_2/tower/conv_1/batchnorm/beta:0' shape=(64,) dtype=float32>,)
mixed_2/tower/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_2/tower/conv_1/batchnorm/gamma:0' shape=(64,) dtype=float32>,)
mixed_2/tower/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_2/tower/conv_1/batchnorm/moving_mean:0' shape=(64,) dtype=float32>,)
mixed_2/tower/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_2/tower/conv_1/batchnorm/moving_variance:0' shape=(64,) dtype=float32>,)
mixed_2/tower/conv_1/batchnorm
(<tf.Tensor 'mixed_2/tower/conv_1/batchnorm:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower/conv_1/CheckNumerics
(<tf.Tensor 'mixed_2/tower/conv_1/CheckNumerics:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower/conv_1/control_dependency
(<tf.Tensor 'mixed_2/tower/conv_1/control_dependency:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower/conv_1
(<tf.Tensor 'mixed_2/tower/conv_1:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower_1/conv/conv2d_params
(<tf.Tensor 'mixed_2/tower_1/conv/conv2d_params:0' shape=(1, 1, 288, 64) dtype=float32>,)
mixed_2/tower_1/conv/Conv2D
(<tf.Tensor 'mixed_2/tower_1/conv/Conv2D:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower_1/conv/batchnorm/beta
(<tf.Tensor 'mixed_2/tower_1/conv/batchnorm/beta:0' shape=(64,) dtype=float32>,)
mixed_2/tower_1/conv/batchnorm/gamma
(<tf.Tensor 'mixed_2/tower_1/conv/batchnorm/gamma:0' shape=(64,) dtype=float32>,)
mixed_2/tower_1/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_2/tower_1/conv/batchnorm/moving_mean:0' shape=(64,) dtype=float32>,)
mixed_2/tower_1/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_2/tower_1/conv/batchnorm/moving_variance:0' shape=(64,) dtype=float32>,)
mixed_2/tower_1/conv/batchnorm
(<tf.Tensor 'mixed_2/tower_1/conv/batchnorm:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower_1/conv/CheckNumerics
(<tf.Tensor 'mixed_2/tower_1/conv/CheckNumerics:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower_1/conv/control_dependency
(<tf.Tensor 'mixed_2/tower_1/conv/control_dependency:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower_1/conv
(<tf.Tensor 'mixed_2/tower_1/conv:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower_1/conv_1/conv2d_params
(<tf.Tensor 'mixed_2/tower_1/conv_1/conv2d_params:0' shape=(3, 3, 64, 96) dtype=float32>,)
mixed_2/tower_1/conv_1/Conv2D
(<tf.Tensor 'mixed_2/tower_1/conv_1/Conv2D:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_2/tower_1/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_2/tower_1/conv_1/batchnorm/beta:0' shape=(96,) dtype=float32>,)
mixed_2/tower_1/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_2/tower_1/conv_1/batchnorm/gamma:0' shape=(96,) dtype=float32>,)
mixed_2/tower_1/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_2/tower_1/conv_1/batchnorm/moving_mean:0' shape=(96,) dtype=float32>,)
mixed_2/tower_1/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_2/tower_1/conv_1/batchnorm/moving_variance:0' shape=(96,) dtype=float32>,)
mixed_2/tower_1/conv_1/batchnorm
(<tf.Tensor 'mixed_2/tower_1/conv_1/batchnorm:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_2/tower_1/conv_1/CheckNumerics
(<tf.Tensor 'mixed_2/tower_1/conv_1/CheckNumerics:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_2/tower_1/conv_1/control_dependency
(<tf.Tensor 'mixed_2/tower_1/conv_1/control_dependency:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_2/tower_1/conv_1
(<tf.Tensor 'mixed_2/tower_1/conv_1:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_2/tower_1/conv_2/conv2d_params
(<tf.Tensor 'mixed_2/tower_1/conv_2/conv2d_params:0' shape=(3, 3, 96, 96) dtype=float32>,)
mixed_2/tower_1/conv_2/Conv2D
(<tf.Tensor 'mixed_2/tower_1/conv_2/Conv2D:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_2/tower_1/conv_2/batchnorm/beta
(<tf.Tensor 'mixed_2/tower_1/conv_2/batchnorm/beta:0' shape=(96,) dtype=float32>,)
mixed_2/tower_1/conv_2/batchnorm/gamma
(<tf.Tensor 'mixed_2/tower_1/conv_2/batchnorm/gamma:0' shape=(96,) dtype=float32>,)
mixed_2/tower_1/conv_2/batchnorm/moving_mean
(<tf.Tensor 'mixed_2/tower_1/conv_2/batchnorm/moving_mean:0' shape=(96,) dtype=float32>,)
mixed_2/tower_1/conv_2/batchnorm/moving_variance
(<tf.Tensor 'mixed_2/tower_1/conv_2/batchnorm/moving_variance:0' shape=(96,) dtype=float32>,)
mixed_2/tower_1/conv_2/batchnorm
(<tf.Tensor 'mixed_2/tower_1/conv_2/batchnorm:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_2/tower_1/conv_2/CheckNumerics
(<tf.Tensor 'mixed_2/tower_1/conv_2/CheckNumerics:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_2/tower_1/conv_2/control_dependency
(<tf.Tensor 'mixed_2/tower_1/conv_2/control_dependency:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_2/tower_1/conv_2
(<tf.Tensor 'mixed_2/tower_1/conv_2:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_2/tower_2/pool
(<tf.Tensor 'mixed_2/tower_2/pool:0' shape=(1, 35, 35, 288) dtype=float32>,)
mixed_2/tower_2/conv/conv2d_params
(<tf.Tensor 'mixed_2/tower_2/conv/conv2d_params:0' shape=(1, 1, 288, 64) dtype=float32>,)
mixed_2/tower_2/conv/Conv2D
(<tf.Tensor 'mixed_2/tower_2/conv/Conv2D:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower_2/conv/batchnorm/beta
(<tf.Tensor 'mixed_2/tower_2/conv/batchnorm/beta:0' shape=(64,) dtype=float32>,)
mixed_2/tower_2/conv/batchnorm/gamma
(<tf.Tensor 'mixed_2/tower_2/conv/batchnorm/gamma:0' shape=(64,) dtype=float32>,)
mixed_2/tower_2/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_2/tower_2/conv/batchnorm/moving_mean:0' shape=(64,) dtype=float32>,)
mixed_2/tower_2/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_2/tower_2/conv/batchnorm/moving_variance:0' shape=(64,) dtype=float32>,)
mixed_2/tower_2/conv/batchnorm
(<tf.Tensor 'mixed_2/tower_2/conv/batchnorm:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower_2/conv/CheckNumerics
(<tf.Tensor 'mixed_2/tower_2/conv/CheckNumerics:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower_2/conv/control_dependency
(<tf.Tensor 'mixed_2/tower_2/conv/control_dependency:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/tower_2/conv
(<tf.Tensor 'mixed_2/tower_2/conv:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_2/join/concat_dim
(<tf.Tensor 'mixed_2/join/concat_dim:0' shape=() dtype=int32>,)
mixed_2/join
(<tf.Tensor 'mixed_2/join:0' shape=(1, 35, 35, 288) dtype=float32>,)
mixed_3/conv/conv2d_params
(<tf.Tensor 'mixed_3/conv/conv2d_params:0' shape=(3, 3, 288, 384) dtype=float32>,)
mixed_3/conv/Conv2D
(<tf.Tensor 'mixed_3/conv/Conv2D:0' shape=(1, 17, 17, 384) dtype=float32>,)
mixed_3/conv/batchnorm/beta
(<tf.Tensor 'mixed_3/conv/batchnorm/beta:0' shape=(384,) dtype=float32>,)
mixed_3/conv/batchnorm/gamma
(<tf.Tensor 'mixed_3/conv/batchnorm/gamma:0' shape=(384,) dtype=float32>,)
mixed_3/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_3/conv/batchnorm/moving_mean:0' shape=(384,) dtype=float32>,)
mixed_3/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_3/conv/batchnorm/moving_variance:0' shape=(384,) dtype=float32>,)
mixed_3/conv/batchnorm
(<tf.Tensor 'mixed_3/conv/batchnorm:0' shape=(1, 17, 17, 384) dtype=float32>,)
mixed_3/conv/CheckNumerics
(<tf.Tensor 'mixed_3/conv/CheckNumerics:0' shape=(1, 17, 17, 384) dtype=float32>,)
mixed_3/conv/control_dependency
(<tf.Tensor 'mixed_3/conv/control_dependency:0' shape=(1, 17, 17, 384) dtype=float32>,)
mixed_3/conv
(<tf.Tensor 'mixed_3/conv:0' shape=(1, 17, 17, 384) dtype=float32>,)
mixed_3/tower/conv/conv2d_params
(<tf.Tensor 'mixed_3/tower/conv/conv2d_params:0' shape=(1, 1, 288, 64) dtype=float32>,)
mixed_3/tower/conv/Conv2D
(<tf.Tensor 'mixed_3/tower/conv/Conv2D:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_3/tower/conv/batchnorm/beta
(<tf.Tensor 'mixed_3/tower/conv/batchnorm/beta:0' shape=(64,) dtype=float32>,)
mixed_3/tower/conv/batchnorm/gamma
(<tf.Tensor 'mixed_3/tower/conv/batchnorm/gamma:0' shape=(64,) dtype=float32>,)
mixed_3/tower/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_3/tower/conv/batchnorm/moving_mean:0' shape=(64,) dtype=float32>,)
mixed_3/tower/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_3/tower/conv/batchnorm/moving_variance:0' shape=(64,) dtype=float32>,)
mixed_3/tower/conv/batchnorm
(<tf.Tensor 'mixed_3/tower/conv/batchnorm:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_3/tower/conv/CheckNumerics
(<tf.Tensor 'mixed_3/tower/conv/CheckNumerics:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_3/tower/conv/control_dependency
(<tf.Tensor 'mixed_3/tower/conv/control_dependency:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_3/tower/conv
(<tf.Tensor 'mixed_3/tower/conv:0' shape=(1, 35, 35, 64) dtype=float32>,)
mixed_3/tower/conv_1/conv2d_params
(<tf.Tensor 'mixed_3/tower/conv_1/conv2d_params:0' shape=(3, 3, 64, 96) dtype=float32>,)
mixed_3/tower/conv_1/Conv2D
(<tf.Tensor 'mixed_3/tower/conv_1/Conv2D:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_3/tower/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_3/tower/conv_1/batchnorm/beta:0' shape=(96,) dtype=float32>,)
mixed_3/tower/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_3/tower/conv_1/batchnorm/gamma:0' shape=(96,) dtype=float32>,)
mixed_3/tower/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_3/tower/conv_1/batchnorm/moving_mean:0' shape=(96,) dtype=float32>,)
mixed_3/tower/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_3/tower/conv_1/batchnorm/moving_variance:0' shape=(96,) dtype=float32>,)
mixed_3/tower/conv_1/batchnorm
(<tf.Tensor 'mixed_3/tower/conv_1/batchnorm:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_3/tower/conv_1/CheckNumerics
(<tf.Tensor 'mixed_3/tower/conv_1/CheckNumerics:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_3/tower/conv_1/control_dependency
(<tf.Tensor 'mixed_3/tower/conv_1/control_dependency:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_3/tower/conv_1
(<tf.Tensor 'mixed_3/tower/conv_1:0' shape=(1, 35, 35, 96) dtype=float32>,)
mixed_3/tower/conv_2/conv2d_params
(<tf.Tensor 'mixed_3/tower/conv_2/conv2d_params:0' shape=(3, 3, 96, 96) dtype=float32>,)
mixed_3/tower/conv_2/Conv2D
(<tf.Tensor 'mixed_3/tower/conv_2/Conv2D:0' shape=(1, 17, 17, 96) dtype=float32>,)
mixed_3/tower/conv_2/batchnorm/beta
(<tf.Tensor 'mixed_3/tower/conv_2/batchnorm/beta:0' shape=(96,) dtype=float32>,)
mixed_3/tower/conv_2/batchnorm/gamma
(<tf.Tensor 'mixed_3/tower/conv_2/batchnorm/gamma:0' shape=(96,) dtype=float32>,)
mixed_3/tower/conv_2/batchnorm/moving_mean
(<tf.Tensor 'mixed_3/tower/conv_2/batchnorm/moving_mean:0' shape=(96,) dtype=float32>,)
mixed_3/tower/conv_2/batchnorm/moving_variance
(<tf.Tensor 'mixed_3/tower/conv_2/batchnorm/moving_variance:0' shape=(96,) dtype=float32>,)
mixed_3/tower/conv_2/batchnorm
(<tf.Tensor 'mixed_3/tower/conv_2/batchnorm:0' shape=(1, 17, 17, 96) dtype=float32>,)
mixed_3/tower/conv_2/CheckNumerics
(<tf.Tensor 'mixed_3/tower/conv_2/CheckNumerics:0' shape=(1, 17, 17, 96) dtype=float32>,)
mixed_3/tower/conv_2/control_dependency
(<tf.Tensor 'mixed_3/tower/conv_2/control_dependency:0' shape=(1, 17, 17, 96) dtype=float32>,)
mixed_3/tower/conv_2
(<tf.Tensor 'mixed_3/tower/conv_2:0' shape=(1, 17, 17, 96) dtype=float32>,)
mixed_3/pool/CheckNumerics
(<tf.Tensor 'mixed_3/pool/CheckNumerics:0' shape=(1, 35, 35, 288) dtype=float32>,)
mixed_3/pool/control_dependency
(<tf.Tensor 'mixed_3/pool/control_dependency:0' shape=(1, 35, 35, 288) dtype=float32>,)
mixed_3/pool
(<tf.Tensor 'mixed_3/pool:0' shape=(1, 17, 17, 288) dtype=float32>,)
mixed_3/join/concat_dim
(<tf.Tensor 'mixed_3/join/concat_dim:0' shape=() dtype=int32>,)
mixed_3/join
(<tf.Tensor 'mixed_3/join:0' shape=(1, 17, 17, 768) dtype=float32>,)
mixed_4/conv/conv2d_params
(<tf.Tensor 'mixed_4/conv/conv2d_params:0' shape=(1, 1, 768, 192) dtype=float32>,)
mixed_4/conv/Conv2D
(<tf.Tensor 'mixed_4/conv/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/conv/batchnorm/beta
(<tf.Tensor 'mixed_4/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_4/conv/batchnorm/gamma
(<tf.Tensor 'mixed_4/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_4/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_4/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_4/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_4/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_4/conv/batchnorm
(<tf.Tensor 'mixed_4/conv/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/conv/CheckNumerics
(<tf.Tensor 'mixed_4/conv/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/conv/control_dependency
(<tf.Tensor 'mixed_4/conv/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/conv
(<tf.Tensor 'mixed_4/conv:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower/conv/conv2d_params
(<tf.Tensor 'mixed_4/tower/conv/conv2d_params:0' shape=(1, 1, 768, 128) dtype=float32>,)
mixed_4/tower/conv/Conv2D
(<tf.Tensor 'mixed_4/tower/conv/Conv2D:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower/conv/batchnorm/beta
(<tf.Tensor 'mixed_4/tower/conv/batchnorm/beta:0' shape=(128,) dtype=float32>,)
mixed_4/tower/conv/batchnorm/gamma
(<tf.Tensor 'mixed_4/tower/conv/batchnorm/gamma:0' shape=(128,) dtype=float32>,)
mixed_4/tower/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_4/tower/conv/batchnorm/moving_mean:0' shape=(128,) dtype=float32>,)
mixed_4/tower/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_4/tower/conv/batchnorm/moving_variance:0' shape=(128,) dtype=float32>,)
mixed_4/tower/conv/batchnorm
(<tf.Tensor 'mixed_4/tower/conv/batchnorm:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower/conv/CheckNumerics
(<tf.Tensor 'mixed_4/tower/conv/CheckNumerics:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower/conv/control_dependency
(<tf.Tensor 'mixed_4/tower/conv/control_dependency:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower/conv
(<tf.Tensor 'mixed_4/tower/conv:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower/conv_1/conv2d_params
(<tf.Tensor 'mixed_4/tower/conv_1/conv2d_params:0' shape=(1, 7, 128, 128) dtype=float32>,)
mixed_4/tower/conv_1/Conv2D
(<tf.Tensor 'mixed_4/tower/conv_1/Conv2D:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_4/tower/conv_1/batchnorm/beta:0' shape=(128,) dtype=float32>,)
mixed_4/tower/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_4/tower/conv_1/batchnorm/gamma:0' shape=(128,) dtype=float32>,)
mixed_4/tower/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_4/tower/conv_1/batchnorm/moving_mean:0' shape=(128,) dtype=float32>,)
mixed_4/tower/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_4/tower/conv_1/batchnorm/moving_variance:0' shape=(128,) dtype=float32>,)
mixed_4/tower/conv_1/batchnorm
(<tf.Tensor 'mixed_4/tower/conv_1/batchnorm:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower/conv_1/CheckNumerics
(<tf.Tensor 'mixed_4/tower/conv_1/CheckNumerics:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower/conv_1/control_dependency
(<tf.Tensor 'mixed_4/tower/conv_1/control_dependency:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower/conv_1
(<tf.Tensor 'mixed_4/tower/conv_1:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower/conv_2/conv2d_params
(<tf.Tensor 'mixed_4/tower/conv_2/conv2d_params:0' shape=(7, 1, 128, 192) dtype=float32>,)
mixed_4/tower/conv_2/Conv2D
(<tf.Tensor 'mixed_4/tower/conv_2/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower/conv_2/batchnorm/beta
(<tf.Tensor 'mixed_4/tower/conv_2/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_4/tower/conv_2/batchnorm/gamma
(<tf.Tensor 'mixed_4/tower/conv_2/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_4/tower/conv_2/batchnorm/moving_mean
(<tf.Tensor 'mixed_4/tower/conv_2/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_4/tower/conv_2/batchnorm/moving_variance
(<tf.Tensor 'mixed_4/tower/conv_2/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_4/tower/conv_2/batchnorm
(<tf.Tensor 'mixed_4/tower/conv_2/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower/conv_2/CheckNumerics
(<tf.Tensor 'mixed_4/tower/conv_2/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower/conv_2/control_dependency
(<tf.Tensor 'mixed_4/tower/conv_2/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower/conv_2
(<tf.Tensor 'mixed_4/tower/conv_2:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower_1/conv/conv2d_params
(<tf.Tensor 'mixed_4/tower_1/conv/conv2d_params:0' shape=(1, 1, 768, 128) dtype=float32>,)
mixed_4/tower_1/conv/Conv2D
(<tf.Tensor 'mixed_4/tower_1/conv/Conv2D:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv/batchnorm/beta
(<tf.Tensor 'mixed_4/tower_1/conv/batchnorm/beta:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv/batchnorm/gamma
(<tf.Tensor 'mixed_4/tower_1/conv/batchnorm/gamma:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_4/tower_1/conv/batchnorm/moving_mean:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_4/tower_1/conv/batchnorm/moving_variance:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv/batchnorm
(<tf.Tensor 'mixed_4/tower_1/conv/batchnorm:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv/CheckNumerics
(<tf.Tensor 'mixed_4/tower_1/conv/CheckNumerics:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv/control_dependency
(<tf.Tensor 'mixed_4/tower_1/conv/control_dependency:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv
(<tf.Tensor 'mixed_4/tower_1/conv:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_1/conv2d_params
(<tf.Tensor 'mixed_4/tower_1/conv_1/conv2d_params:0' shape=(7, 1, 128, 128) dtype=float32>,)
mixed_4/tower_1/conv_1/Conv2D
(<tf.Tensor 'mixed_4/tower_1/conv_1/Conv2D:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_4/tower_1/conv_1/batchnorm/beta:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_4/tower_1/conv_1/batchnorm/gamma:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_4/tower_1/conv_1/batchnorm/moving_mean:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_4/tower_1/conv_1/batchnorm/moving_variance:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv_1/batchnorm
(<tf.Tensor 'mixed_4/tower_1/conv_1/batchnorm:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_1/CheckNumerics
(<tf.Tensor 'mixed_4/tower_1/conv_1/CheckNumerics:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_1/control_dependency
(<tf.Tensor 'mixed_4/tower_1/conv_1/control_dependency:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_1
(<tf.Tensor 'mixed_4/tower_1/conv_1:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_2/conv2d_params
(<tf.Tensor 'mixed_4/tower_1/conv_2/conv2d_params:0' shape=(1, 7, 128, 128) dtype=float32>,)
mixed_4/tower_1/conv_2/Conv2D
(<tf.Tensor 'mixed_4/tower_1/conv_2/Conv2D:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_2/batchnorm/beta
(<tf.Tensor 'mixed_4/tower_1/conv_2/batchnorm/beta:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv_2/batchnorm/gamma
(<tf.Tensor 'mixed_4/tower_1/conv_2/batchnorm/gamma:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv_2/batchnorm/moving_mean
(<tf.Tensor 'mixed_4/tower_1/conv_2/batchnorm/moving_mean:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv_2/batchnorm/moving_variance
(<tf.Tensor 'mixed_4/tower_1/conv_2/batchnorm/moving_variance:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv_2/batchnorm
(<tf.Tensor 'mixed_4/tower_1/conv_2/batchnorm:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_2/CheckNumerics
(<tf.Tensor 'mixed_4/tower_1/conv_2/CheckNumerics:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_2/control_dependency
(<tf.Tensor 'mixed_4/tower_1/conv_2/control_dependency:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_2
(<tf.Tensor 'mixed_4/tower_1/conv_2:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_3/conv2d_params
(<tf.Tensor 'mixed_4/tower_1/conv_3/conv2d_params:0' shape=(7, 1, 128, 128) dtype=float32>,)
mixed_4/tower_1/conv_3/Conv2D
(<tf.Tensor 'mixed_4/tower_1/conv_3/Conv2D:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_3/batchnorm/beta
(<tf.Tensor 'mixed_4/tower_1/conv_3/batchnorm/beta:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv_3/batchnorm/gamma
(<tf.Tensor 'mixed_4/tower_1/conv_3/batchnorm/gamma:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv_3/batchnorm/moving_mean
(<tf.Tensor 'mixed_4/tower_1/conv_3/batchnorm/moving_mean:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv_3/batchnorm/moving_variance
(<tf.Tensor 'mixed_4/tower_1/conv_3/batchnorm/moving_variance:0' shape=(128,) dtype=float32>,)
mixed_4/tower_1/conv_3/batchnorm
(<tf.Tensor 'mixed_4/tower_1/conv_3/batchnorm:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_3/CheckNumerics
(<tf.Tensor 'mixed_4/tower_1/conv_3/CheckNumerics:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_3/control_dependency
(<tf.Tensor 'mixed_4/tower_1/conv_3/control_dependency:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_3
(<tf.Tensor 'mixed_4/tower_1/conv_3:0' shape=(1, 17, 17, 128) dtype=float32>,)
mixed_4/tower_1/conv_4/conv2d_params
(<tf.Tensor 'mixed_4/tower_1/conv_4/conv2d_params:0' shape=(1, 7, 128, 192) dtype=float32>,)
mixed_4/tower_1/conv_4/Conv2D
(<tf.Tensor 'mixed_4/tower_1/conv_4/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower_1/conv_4/batchnorm/beta
(<tf.Tensor 'mixed_4/tower_1/conv_4/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_4/tower_1/conv_4/batchnorm/gamma
(<tf.Tensor 'mixed_4/tower_1/conv_4/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_4/tower_1/conv_4/batchnorm/moving_mean
(<tf.Tensor 'mixed_4/tower_1/conv_4/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_4/tower_1/conv_4/batchnorm/moving_variance
(<tf.Tensor 'mixed_4/tower_1/conv_4/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_4/tower_1/conv_4/batchnorm
(<tf.Tensor 'mixed_4/tower_1/conv_4/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower_1/conv_4/CheckNumerics
(<tf.Tensor 'mixed_4/tower_1/conv_4/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower_1/conv_4/control_dependency
(<tf.Tensor 'mixed_4/tower_1/conv_4/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower_1/conv_4
(<tf.Tensor 'mixed_4/tower_1/conv_4:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower_2/pool
(<tf.Tensor 'mixed_4/tower_2/pool:0' shape=(1, 17, 17, 768) dtype=float32>,)
mixed_4/tower_2/conv/conv2d_params
(<tf.Tensor 'mixed_4/tower_2/conv/conv2d_params:0' shape=(1, 1, 768, 192) dtype=float32>,)
mixed_4/tower_2/conv/Conv2D
(<tf.Tensor 'mixed_4/tower_2/conv/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower_2/conv/batchnorm/beta
(<tf.Tensor 'mixed_4/tower_2/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_4/tower_2/conv/batchnorm/gamma
(<tf.Tensor 'mixed_4/tower_2/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_4/tower_2/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_4/tower_2/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_4/tower_2/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_4/tower_2/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_4/tower_2/conv/batchnorm
(<tf.Tensor 'mixed_4/tower_2/conv/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower_2/conv/CheckNumerics
(<tf.Tensor 'mixed_4/tower_2/conv/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower_2/conv/control_dependency
(<tf.Tensor 'mixed_4/tower_2/conv/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/tower_2/conv
(<tf.Tensor 'mixed_4/tower_2/conv:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_4/join/concat_dim
(<tf.Tensor 'mixed_4/join/concat_dim:0' shape=() dtype=int32>,)
mixed_4/join
(<tf.Tensor 'mixed_4/join:0' shape=(1, 17, 17, 768) dtype=float32>,)
mixed_5/conv/conv2d_params
(<tf.Tensor 'mixed_5/conv/conv2d_params:0' shape=(1, 1, 768, 192) dtype=float32>,)
mixed_5/conv/Conv2D
(<tf.Tensor 'mixed_5/conv/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/conv/batchnorm/beta
(<tf.Tensor 'mixed_5/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_5/conv/batchnorm/gamma
(<tf.Tensor 'mixed_5/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_5/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_5/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_5/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_5/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_5/conv/batchnorm
(<tf.Tensor 'mixed_5/conv/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/conv/CheckNumerics
(<tf.Tensor 'mixed_5/conv/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/conv/control_dependency
(<tf.Tensor 'mixed_5/conv/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/conv
(<tf.Tensor 'mixed_5/conv:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower/conv/conv2d_params
(<tf.Tensor 'mixed_5/tower/conv/conv2d_params:0' shape=(1, 1, 768, 160) dtype=float32>,)
mixed_5/tower/conv/Conv2D
(<tf.Tensor 'mixed_5/tower/conv/Conv2D:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower/conv/batchnorm/beta
(<tf.Tensor 'mixed_5/tower/conv/batchnorm/beta:0' shape=(160,) dtype=float32>,)
mixed_5/tower/conv/batchnorm/gamma
(<tf.Tensor 'mixed_5/tower/conv/batchnorm/gamma:0' shape=(160,) dtype=float32>,)
mixed_5/tower/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_5/tower/conv/batchnorm/moving_mean:0' shape=(160,) dtype=float32>,)
mixed_5/tower/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_5/tower/conv/batchnorm/moving_variance:0' shape=(160,) dtype=float32>,)
mixed_5/tower/conv/batchnorm
(<tf.Tensor 'mixed_5/tower/conv/batchnorm:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower/conv/CheckNumerics
(<tf.Tensor 'mixed_5/tower/conv/CheckNumerics:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower/conv/control_dependency
(<tf.Tensor 'mixed_5/tower/conv/control_dependency:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower/conv
(<tf.Tensor 'mixed_5/tower/conv:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower/conv_1/conv2d_params
(<tf.Tensor 'mixed_5/tower/conv_1/conv2d_params:0' shape=(1, 7, 160, 160) dtype=float32>,)
mixed_5/tower/conv_1/Conv2D
(<tf.Tensor 'mixed_5/tower/conv_1/Conv2D:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_5/tower/conv_1/batchnorm/beta:0' shape=(160,) dtype=float32>,)
mixed_5/tower/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_5/tower/conv_1/batchnorm/gamma:0' shape=(160,) dtype=float32>,)
mixed_5/tower/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_5/tower/conv_1/batchnorm/moving_mean:0' shape=(160,) dtype=float32>,)
mixed_5/tower/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_5/tower/conv_1/batchnorm/moving_variance:0' shape=(160,) dtype=float32>,)
mixed_5/tower/conv_1/batchnorm
(<tf.Tensor 'mixed_5/tower/conv_1/batchnorm:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower/conv_1/CheckNumerics
(<tf.Tensor 'mixed_5/tower/conv_1/CheckNumerics:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower/conv_1/control_dependency
(<tf.Tensor 'mixed_5/tower/conv_1/control_dependency:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower/conv_1
(<tf.Tensor 'mixed_5/tower/conv_1:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower/conv_2/conv2d_params
(<tf.Tensor 'mixed_5/tower/conv_2/conv2d_params:0' shape=(7, 1, 160, 192) dtype=float32>,)
mixed_5/tower/conv_2/Conv2D
(<tf.Tensor 'mixed_5/tower/conv_2/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower/conv_2/batchnorm/beta
(<tf.Tensor 'mixed_5/tower/conv_2/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_5/tower/conv_2/batchnorm/gamma
(<tf.Tensor 'mixed_5/tower/conv_2/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_5/tower/conv_2/batchnorm/moving_mean
(<tf.Tensor 'mixed_5/tower/conv_2/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_5/tower/conv_2/batchnorm/moving_variance
(<tf.Tensor 'mixed_5/tower/conv_2/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_5/tower/conv_2/batchnorm
(<tf.Tensor 'mixed_5/tower/conv_2/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower/conv_2/CheckNumerics
(<tf.Tensor 'mixed_5/tower/conv_2/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower/conv_2/control_dependency
(<tf.Tensor 'mixed_5/tower/conv_2/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower/conv_2
(<tf.Tensor 'mixed_5/tower/conv_2:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower_1/conv/conv2d_params
(<tf.Tensor 'mixed_5/tower_1/conv/conv2d_params:0' shape=(1, 1, 768, 160) dtype=float32>,)
mixed_5/tower_1/conv/Conv2D
(<tf.Tensor 'mixed_5/tower_1/conv/Conv2D:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv/batchnorm/beta
(<tf.Tensor 'mixed_5/tower_1/conv/batchnorm/beta:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv/batchnorm/gamma
(<tf.Tensor 'mixed_5/tower_1/conv/batchnorm/gamma:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_5/tower_1/conv/batchnorm/moving_mean:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_5/tower_1/conv/batchnorm/moving_variance:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv/batchnorm
(<tf.Tensor 'mixed_5/tower_1/conv/batchnorm:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv/CheckNumerics
(<tf.Tensor 'mixed_5/tower_1/conv/CheckNumerics:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv/control_dependency
(<tf.Tensor 'mixed_5/tower_1/conv/control_dependency:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv
(<tf.Tensor 'mixed_5/tower_1/conv:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_1/conv2d_params
(<tf.Tensor 'mixed_5/tower_1/conv_1/conv2d_params:0' shape=(7, 1, 160, 160) dtype=float32>,)
mixed_5/tower_1/conv_1/Conv2D
(<tf.Tensor 'mixed_5/tower_1/conv_1/Conv2D:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_5/tower_1/conv_1/batchnorm/beta:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_5/tower_1/conv_1/batchnorm/gamma:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_5/tower_1/conv_1/batchnorm/moving_mean:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_5/tower_1/conv_1/batchnorm/moving_variance:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv_1/batchnorm
(<tf.Tensor 'mixed_5/tower_1/conv_1/batchnorm:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_1/CheckNumerics
(<tf.Tensor 'mixed_5/tower_1/conv_1/CheckNumerics:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_1/control_dependency
(<tf.Tensor 'mixed_5/tower_1/conv_1/control_dependency:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_1
(<tf.Tensor 'mixed_5/tower_1/conv_1:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_2/conv2d_params
(<tf.Tensor 'mixed_5/tower_1/conv_2/conv2d_params:0' shape=(1, 7, 160, 160) dtype=float32>,)
mixed_5/tower_1/conv_2/Conv2D
(<tf.Tensor 'mixed_5/tower_1/conv_2/Conv2D:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_2/batchnorm/beta
(<tf.Tensor 'mixed_5/tower_1/conv_2/batchnorm/beta:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv_2/batchnorm/gamma
(<tf.Tensor 'mixed_5/tower_1/conv_2/batchnorm/gamma:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv_2/batchnorm/moving_mean
(<tf.Tensor 'mixed_5/tower_1/conv_2/batchnorm/moving_mean:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv_2/batchnorm/moving_variance
(<tf.Tensor 'mixed_5/tower_1/conv_2/batchnorm/moving_variance:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv_2/batchnorm
(<tf.Tensor 'mixed_5/tower_1/conv_2/batchnorm:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_2/CheckNumerics
(<tf.Tensor 'mixed_5/tower_1/conv_2/CheckNumerics:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_2/control_dependency
(<tf.Tensor 'mixed_5/tower_1/conv_2/control_dependency:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_2
(<tf.Tensor 'mixed_5/tower_1/conv_2:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_3/conv2d_params
(<tf.Tensor 'mixed_5/tower_1/conv_3/conv2d_params:0' shape=(7, 1, 160, 160) dtype=float32>,)
mixed_5/tower_1/conv_3/Conv2D
(<tf.Tensor 'mixed_5/tower_1/conv_3/Conv2D:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_3/batchnorm/beta
(<tf.Tensor 'mixed_5/tower_1/conv_3/batchnorm/beta:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv_3/batchnorm/gamma
(<tf.Tensor 'mixed_5/tower_1/conv_3/batchnorm/gamma:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv_3/batchnorm/moving_mean
(<tf.Tensor 'mixed_5/tower_1/conv_3/batchnorm/moving_mean:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv_3/batchnorm/moving_variance
(<tf.Tensor 'mixed_5/tower_1/conv_3/batchnorm/moving_variance:0' shape=(160,) dtype=float32>,)
mixed_5/tower_1/conv_3/batchnorm
(<tf.Tensor 'mixed_5/tower_1/conv_3/batchnorm:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_3/CheckNumerics
(<tf.Tensor 'mixed_5/tower_1/conv_3/CheckNumerics:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_3/control_dependency
(<tf.Tensor 'mixed_5/tower_1/conv_3/control_dependency:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_3
(<tf.Tensor 'mixed_5/tower_1/conv_3:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_5/tower_1/conv_4/conv2d_params
(<tf.Tensor 'mixed_5/tower_1/conv_4/conv2d_params:0' shape=(1, 7, 160, 192) dtype=float32>,)
mixed_5/tower_1/conv_4/Conv2D
(<tf.Tensor 'mixed_5/tower_1/conv_4/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower_1/conv_4/batchnorm/beta
(<tf.Tensor 'mixed_5/tower_1/conv_4/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_5/tower_1/conv_4/batchnorm/gamma
(<tf.Tensor 'mixed_5/tower_1/conv_4/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_5/tower_1/conv_4/batchnorm/moving_mean
(<tf.Tensor 'mixed_5/tower_1/conv_4/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_5/tower_1/conv_4/batchnorm/moving_variance
(<tf.Tensor 'mixed_5/tower_1/conv_4/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_5/tower_1/conv_4/batchnorm
(<tf.Tensor 'mixed_5/tower_1/conv_4/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower_1/conv_4/CheckNumerics
(<tf.Tensor 'mixed_5/tower_1/conv_4/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower_1/conv_4/control_dependency
(<tf.Tensor 'mixed_5/tower_1/conv_4/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower_1/conv_4
(<tf.Tensor 'mixed_5/tower_1/conv_4:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower_2/pool
(<tf.Tensor 'mixed_5/tower_2/pool:0' shape=(1, 17, 17, 768) dtype=float32>,)
mixed_5/tower_2/conv/conv2d_params
(<tf.Tensor 'mixed_5/tower_2/conv/conv2d_params:0' shape=(1, 1, 768, 192) dtype=float32>,)
mixed_5/tower_2/conv/Conv2D
(<tf.Tensor 'mixed_5/tower_2/conv/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower_2/conv/batchnorm/beta
(<tf.Tensor 'mixed_5/tower_2/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_5/tower_2/conv/batchnorm/gamma
(<tf.Tensor 'mixed_5/tower_2/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_5/tower_2/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_5/tower_2/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_5/tower_2/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_5/tower_2/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_5/tower_2/conv/batchnorm
(<tf.Tensor 'mixed_5/tower_2/conv/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower_2/conv/CheckNumerics
(<tf.Tensor 'mixed_5/tower_2/conv/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower_2/conv/control_dependency
(<tf.Tensor 'mixed_5/tower_2/conv/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/tower_2/conv
(<tf.Tensor 'mixed_5/tower_2/conv:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_5/join/concat_dim
(<tf.Tensor 'mixed_5/join/concat_dim:0' shape=() dtype=int32>,)
mixed_5/join
(<tf.Tensor 'mixed_5/join:0' shape=(1, 17, 17, 768) dtype=float32>,)
mixed_6/conv/conv2d_params
(<tf.Tensor 'mixed_6/conv/conv2d_params:0' shape=(1, 1, 768, 192) dtype=float32>,)
mixed_6/conv/Conv2D
(<tf.Tensor 'mixed_6/conv/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/conv/batchnorm/beta
(<tf.Tensor 'mixed_6/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_6/conv/batchnorm/gamma
(<tf.Tensor 'mixed_6/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_6/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_6/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_6/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_6/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_6/conv/batchnorm
(<tf.Tensor 'mixed_6/conv/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/conv/CheckNumerics
(<tf.Tensor 'mixed_6/conv/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/conv/control_dependency
(<tf.Tensor 'mixed_6/conv/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/conv
(<tf.Tensor 'mixed_6/conv:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower/conv/conv2d_params
(<tf.Tensor 'mixed_6/tower/conv/conv2d_params:0' shape=(1, 1, 768, 160) dtype=float32>,)
mixed_6/tower/conv/Conv2D
(<tf.Tensor 'mixed_6/tower/conv/Conv2D:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower/conv/batchnorm/beta
(<tf.Tensor 'mixed_6/tower/conv/batchnorm/beta:0' shape=(160,) dtype=float32>,)
mixed_6/tower/conv/batchnorm/gamma
(<tf.Tensor 'mixed_6/tower/conv/batchnorm/gamma:0' shape=(160,) dtype=float32>,)
mixed_6/tower/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_6/tower/conv/batchnorm/moving_mean:0' shape=(160,) dtype=float32>,)
mixed_6/tower/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_6/tower/conv/batchnorm/moving_variance:0' shape=(160,) dtype=float32>,)
mixed_6/tower/conv/batchnorm
(<tf.Tensor 'mixed_6/tower/conv/batchnorm:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower/conv/CheckNumerics
(<tf.Tensor 'mixed_6/tower/conv/CheckNumerics:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower/conv/control_dependency
(<tf.Tensor 'mixed_6/tower/conv/control_dependency:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower/conv
(<tf.Tensor 'mixed_6/tower/conv:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower/conv_1/conv2d_params
(<tf.Tensor 'mixed_6/tower/conv_1/conv2d_params:0' shape=(1, 7, 160, 160) dtype=float32>,)
mixed_6/tower/conv_1/Conv2D
(<tf.Tensor 'mixed_6/tower/conv_1/Conv2D:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_6/tower/conv_1/batchnorm/beta:0' shape=(160,) dtype=float32>,)
mixed_6/tower/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_6/tower/conv_1/batchnorm/gamma:0' shape=(160,) dtype=float32>,)
mixed_6/tower/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_6/tower/conv_1/batchnorm/moving_mean:0' shape=(160,) dtype=float32>,)
mixed_6/tower/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_6/tower/conv_1/batchnorm/moving_variance:0' shape=(160,) dtype=float32>,)
mixed_6/tower/conv_1/batchnorm
(<tf.Tensor 'mixed_6/tower/conv_1/batchnorm:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower/conv_1/CheckNumerics
(<tf.Tensor 'mixed_6/tower/conv_1/CheckNumerics:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower/conv_1/control_dependency
(<tf.Tensor 'mixed_6/tower/conv_1/control_dependency:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower/conv_1
(<tf.Tensor 'mixed_6/tower/conv_1:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower/conv_2/conv2d_params
(<tf.Tensor 'mixed_6/tower/conv_2/conv2d_params:0' shape=(7, 1, 160, 192) dtype=float32>,)
mixed_6/tower/conv_2/Conv2D
(<tf.Tensor 'mixed_6/tower/conv_2/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower/conv_2/batchnorm/beta
(<tf.Tensor 'mixed_6/tower/conv_2/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_6/tower/conv_2/batchnorm/gamma
(<tf.Tensor 'mixed_6/tower/conv_2/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_6/tower/conv_2/batchnorm/moving_mean
(<tf.Tensor 'mixed_6/tower/conv_2/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_6/tower/conv_2/batchnorm/moving_variance
(<tf.Tensor 'mixed_6/tower/conv_2/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_6/tower/conv_2/batchnorm
(<tf.Tensor 'mixed_6/tower/conv_2/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower/conv_2/CheckNumerics
(<tf.Tensor 'mixed_6/tower/conv_2/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower/conv_2/control_dependency
(<tf.Tensor 'mixed_6/tower/conv_2/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower/conv_2
(<tf.Tensor 'mixed_6/tower/conv_2:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower_1/conv/conv2d_params
(<tf.Tensor 'mixed_6/tower_1/conv/conv2d_params:0' shape=(1, 1, 768, 160) dtype=float32>,)
mixed_6/tower_1/conv/Conv2D
(<tf.Tensor 'mixed_6/tower_1/conv/Conv2D:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv/batchnorm/beta
(<tf.Tensor 'mixed_6/tower_1/conv/batchnorm/beta:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv/batchnorm/gamma
(<tf.Tensor 'mixed_6/tower_1/conv/batchnorm/gamma:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_6/tower_1/conv/batchnorm/moving_mean:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_6/tower_1/conv/batchnorm/moving_variance:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv/batchnorm
(<tf.Tensor 'mixed_6/tower_1/conv/batchnorm:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv/CheckNumerics
(<tf.Tensor 'mixed_6/tower_1/conv/CheckNumerics:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv/control_dependency
(<tf.Tensor 'mixed_6/tower_1/conv/control_dependency:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv
(<tf.Tensor 'mixed_6/tower_1/conv:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_1/conv2d_params
(<tf.Tensor 'mixed_6/tower_1/conv_1/conv2d_params:0' shape=(7, 1, 160, 160) dtype=float32>,)
mixed_6/tower_1/conv_1/Conv2D
(<tf.Tensor 'mixed_6/tower_1/conv_1/Conv2D:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_6/tower_1/conv_1/batchnorm/beta:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_6/tower_1/conv_1/batchnorm/gamma:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_6/tower_1/conv_1/batchnorm/moving_mean:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_6/tower_1/conv_1/batchnorm/moving_variance:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv_1/batchnorm
(<tf.Tensor 'mixed_6/tower_1/conv_1/batchnorm:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_1/CheckNumerics
(<tf.Tensor 'mixed_6/tower_1/conv_1/CheckNumerics:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_1/control_dependency
(<tf.Tensor 'mixed_6/tower_1/conv_1/control_dependency:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_1
(<tf.Tensor 'mixed_6/tower_1/conv_1:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_2/conv2d_params
(<tf.Tensor 'mixed_6/tower_1/conv_2/conv2d_params:0' shape=(1, 7, 160, 160) dtype=float32>,)
mixed_6/tower_1/conv_2/Conv2D
(<tf.Tensor 'mixed_6/tower_1/conv_2/Conv2D:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_2/batchnorm/beta
(<tf.Tensor 'mixed_6/tower_1/conv_2/batchnorm/beta:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv_2/batchnorm/gamma
(<tf.Tensor 'mixed_6/tower_1/conv_2/batchnorm/gamma:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv_2/batchnorm/moving_mean
(<tf.Tensor 'mixed_6/tower_1/conv_2/batchnorm/moving_mean:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv_2/batchnorm/moving_variance
(<tf.Tensor 'mixed_6/tower_1/conv_2/batchnorm/moving_variance:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv_2/batchnorm
(<tf.Tensor 'mixed_6/tower_1/conv_2/batchnorm:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_2/CheckNumerics
(<tf.Tensor 'mixed_6/tower_1/conv_2/CheckNumerics:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_2/control_dependency
(<tf.Tensor 'mixed_6/tower_1/conv_2/control_dependency:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_2
(<tf.Tensor 'mixed_6/tower_1/conv_2:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_3/conv2d_params
(<tf.Tensor 'mixed_6/tower_1/conv_3/conv2d_params:0' shape=(7, 1, 160, 160) dtype=float32>,)
mixed_6/tower_1/conv_3/Conv2D
(<tf.Tensor 'mixed_6/tower_1/conv_3/Conv2D:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_3/batchnorm/beta
(<tf.Tensor 'mixed_6/tower_1/conv_3/batchnorm/beta:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv_3/batchnorm/gamma
(<tf.Tensor 'mixed_6/tower_1/conv_3/batchnorm/gamma:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv_3/batchnorm/moving_mean
(<tf.Tensor 'mixed_6/tower_1/conv_3/batchnorm/moving_mean:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv_3/batchnorm/moving_variance
(<tf.Tensor 'mixed_6/tower_1/conv_3/batchnorm/moving_variance:0' shape=(160,) dtype=float32>,)
mixed_6/tower_1/conv_3/batchnorm
(<tf.Tensor 'mixed_6/tower_1/conv_3/batchnorm:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_3/CheckNumerics
(<tf.Tensor 'mixed_6/tower_1/conv_3/CheckNumerics:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_3/control_dependency
(<tf.Tensor 'mixed_6/tower_1/conv_3/control_dependency:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_3
(<tf.Tensor 'mixed_6/tower_1/conv_3:0' shape=(1, 17, 17, 160) dtype=float32>,)
mixed_6/tower_1/conv_4/conv2d_params
(<tf.Tensor 'mixed_6/tower_1/conv_4/conv2d_params:0' shape=(1, 7, 160, 192) dtype=float32>,)
mixed_6/tower_1/conv_4/Conv2D
(<tf.Tensor 'mixed_6/tower_1/conv_4/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower_1/conv_4/batchnorm/beta
(<tf.Tensor 'mixed_6/tower_1/conv_4/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_6/tower_1/conv_4/batchnorm/gamma
(<tf.Tensor 'mixed_6/tower_1/conv_4/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_6/tower_1/conv_4/batchnorm/moving_mean
(<tf.Tensor 'mixed_6/tower_1/conv_4/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_6/tower_1/conv_4/batchnorm/moving_variance
(<tf.Tensor 'mixed_6/tower_1/conv_4/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_6/tower_1/conv_4/batchnorm
(<tf.Tensor 'mixed_6/tower_1/conv_4/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower_1/conv_4/CheckNumerics
(<tf.Tensor 'mixed_6/tower_1/conv_4/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower_1/conv_4/control_dependency
(<tf.Tensor 'mixed_6/tower_1/conv_4/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower_1/conv_4
(<tf.Tensor 'mixed_6/tower_1/conv_4:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower_2/pool
(<tf.Tensor 'mixed_6/tower_2/pool:0' shape=(1, 17, 17, 768) dtype=float32>,)
mixed_6/tower_2/conv/conv2d_params
(<tf.Tensor 'mixed_6/tower_2/conv/conv2d_params:0' shape=(1, 1, 768, 192) dtype=float32>,)
mixed_6/tower_2/conv/Conv2D
(<tf.Tensor 'mixed_6/tower_2/conv/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower_2/conv/batchnorm/beta
(<tf.Tensor 'mixed_6/tower_2/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_6/tower_2/conv/batchnorm/gamma
(<tf.Tensor 'mixed_6/tower_2/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_6/tower_2/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_6/tower_2/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_6/tower_2/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_6/tower_2/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_6/tower_2/conv/batchnorm
(<tf.Tensor 'mixed_6/tower_2/conv/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower_2/conv/CheckNumerics
(<tf.Tensor 'mixed_6/tower_2/conv/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower_2/conv/control_dependency
(<tf.Tensor 'mixed_6/tower_2/conv/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/tower_2/conv
(<tf.Tensor 'mixed_6/tower_2/conv:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_6/join/concat_dim
(<tf.Tensor 'mixed_6/join/concat_dim:0' shape=() dtype=int32>,)
mixed_6/join
(<tf.Tensor 'mixed_6/join:0' shape=(1, 17, 17, 768) dtype=float32>,)
mixed_7/conv/conv2d_params
(<tf.Tensor 'mixed_7/conv/conv2d_params:0' shape=(1, 1, 768, 192) dtype=float32>,)
mixed_7/conv/Conv2D
(<tf.Tensor 'mixed_7/conv/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/conv/batchnorm/beta
(<tf.Tensor 'mixed_7/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_7/conv/batchnorm/gamma
(<tf.Tensor 'mixed_7/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_7/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_7/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_7/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_7/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_7/conv/batchnorm
(<tf.Tensor 'mixed_7/conv/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/conv/CheckNumerics
(<tf.Tensor 'mixed_7/conv/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/conv/control_dependency
(<tf.Tensor 'mixed_7/conv/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/conv
(<tf.Tensor 'mixed_7/conv:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv/conv2d_params
(<tf.Tensor 'mixed_7/tower/conv/conv2d_params:0' shape=(1, 1, 768, 192) dtype=float32>,)
mixed_7/tower/conv/Conv2D
(<tf.Tensor 'mixed_7/tower/conv/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv/batchnorm/beta
(<tf.Tensor 'mixed_7/tower/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_7/tower/conv/batchnorm/gamma
(<tf.Tensor 'mixed_7/tower/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_7/tower/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_7/tower/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_7/tower/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_7/tower/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_7/tower/conv/batchnorm
(<tf.Tensor 'mixed_7/tower/conv/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv/CheckNumerics
(<tf.Tensor 'mixed_7/tower/conv/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv/control_dependency
(<tf.Tensor 'mixed_7/tower/conv/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv
(<tf.Tensor 'mixed_7/tower/conv:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv_1/conv2d_params
(<tf.Tensor 'mixed_7/tower/conv_1/conv2d_params:0' shape=(1, 7, 192, 192) dtype=float32>,)
mixed_7/tower/conv_1/Conv2D
(<tf.Tensor 'mixed_7/tower/conv_1/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_7/tower/conv_1/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_7/tower/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_7/tower/conv_1/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_7/tower/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_7/tower/conv_1/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_7/tower/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_7/tower/conv_1/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_7/tower/conv_1/batchnorm
(<tf.Tensor 'mixed_7/tower/conv_1/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv_1/CheckNumerics
(<tf.Tensor 'mixed_7/tower/conv_1/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv_1/control_dependency
(<tf.Tensor 'mixed_7/tower/conv_1/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv_1
(<tf.Tensor 'mixed_7/tower/conv_1:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv_2/conv2d_params
(<tf.Tensor 'mixed_7/tower/conv_2/conv2d_params:0' shape=(7, 1, 192, 192) dtype=float32>,)
mixed_7/tower/conv_2/Conv2D
(<tf.Tensor 'mixed_7/tower/conv_2/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv_2/batchnorm/beta
(<tf.Tensor 'mixed_7/tower/conv_2/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_7/tower/conv_2/batchnorm/gamma
(<tf.Tensor 'mixed_7/tower/conv_2/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_7/tower/conv_2/batchnorm/moving_mean
(<tf.Tensor 'mixed_7/tower/conv_2/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_7/tower/conv_2/batchnorm/moving_variance
(<tf.Tensor 'mixed_7/tower/conv_2/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_7/tower/conv_2/batchnorm
(<tf.Tensor 'mixed_7/tower/conv_2/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv_2/CheckNumerics
(<tf.Tensor 'mixed_7/tower/conv_2/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv_2/control_dependency
(<tf.Tensor 'mixed_7/tower/conv_2/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower/conv_2
(<tf.Tensor 'mixed_7/tower/conv_2:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv/conv2d_params
(<tf.Tensor 'mixed_7/tower_1/conv/conv2d_params:0' shape=(1, 1, 768, 192) dtype=float32>,)
mixed_7/tower_1/conv/Conv2D
(<tf.Tensor 'mixed_7/tower_1/conv/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv/batchnorm/beta
(<tf.Tensor 'mixed_7/tower_1/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv/batchnorm/gamma
(<tf.Tensor 'mixed_7/tower_1/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_7/tower_1/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_7/tower_1/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv/batchnorm
(<tf.Tensor 'mixed_7/tower_1/conv/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv/CheckNumerics
(<tf.Tensor 'mixed_7/tower_1/conv/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv/control_dependency
(<tf.Tensor 'mixed_7/tower_1/conv/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv
(<tf.Tensor 'mixed_7/tower_1/conv:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_1/conv2d_params
(<tf.Tensor 'mixed_7/tower_1/conv_1/conv2d_params:0' shape=(7, 1, 192, 192) dtype=float32>,)
mixed_7/tower_1/conv_1/Conv2D
(<tf.Tensor 'mixed_7/tower_1/conv_1/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_7/tower_1/conv_1/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_7/tower_1/conv_1/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_7/tower_1/conv_1/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_7/tower_1/conv_1/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_1/batchnorm
(<tf.Tensor 'mixed_7/tower_1/conv_1/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_1/CheckNumerics
(<tf.Tensor 'mixed_7/tower_1/conv_1/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_1/control_dependency
(<tf.Tensor 'mixed_7/tower_1/conv_1/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_1
(<tf.Tensor 'mixed_7/tower_1/conv_1:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_2/conv2d_params
(<tf.Tensor 'mixed_7/tower_1/conv_2/conv2d_params:0' shape=(1, 7, 192, 192) dtype=float32>,)
mixed_7/tower_1/conv_2/Conv2D
(<tf.Tensor 'mixed_7/tower_1/conv_2/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_2/batchnorm/beta
(<tf.Tensor 'mixed_7/tower_1/conv_2/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_2/batchnorm/gamma
(<tf.Tensor 'mixed_7/tower_1/conv_2/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_2/batchnorm/moving_mean
(<tf.Tensor 'mixed_7/tower_1/conv_2/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_2/batchnorm/moving_variance
(<tf.Tensor 'mixed_7/tower_1/conv_2/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_2/batchnorm
(<tf.Tensor 'mixed_7/tower_1/conv_2/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_2/CheckNumerics
(<tf.Tensor 'mixed_7/tower_1/conv_2/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_2/control_dependency
(<tf.Tensor 'mixed_7/tower_1/conv_2/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_2
(<tf.Tensor 'mixed_7/tower_1/conv_2:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_3/conv2d_params
(<tf.Tensor 'mixed_7/tower_1/conv_3/conv2d_params:0' shape=(7, 1, 192, 192) dtype=float32>,)
mixed_7/tower_1/conv_3/Conv2D
(<tf.Tensor 'mixed_7/tower_1/conv_3/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_3/batchnorm/beta
(<tf.Tensor 'mixed_7/tower_1/conv_3/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_3/batchnorm/gamma
(<tf.Tensor 'mixed_7/tower_1/conv_3/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_3/batchnorm/moving_mean
(<tf.Tensor 'mixed_7/tower_1/conv_3/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_3/batchnorm/moving_variance
(<tf.Tensor 'mixed_7/tower_1/conv_3/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_3/batchnorm
(<tf.Tensor 'mixed_7/tower_1/conv_3/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_3/CheckNumerics
(<tf.Tensor 'mixed_7/tower_1/conv_3/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_3/control_dependency
(<tf.Tensor 'mixed_7/tower_1/conv_3/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_3
(<tf.Tensor 'mixed_7/tower_1/conv_3:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_4/conv2d_params
(<tf.Tensor 'mixed_7/tower_1/conv_4/conv2d_params:0' shape=(1, 7, 192, 192) dtype=float32>,)
mixed_7/tower_1/conv_4/Conv2D
(<tf.Tensor 'mixed_7/tower_1/conv_4/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_4/batchnorm/beta
(<tf.Tensor 'mixed_7/tower_1/conv_4/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_4/batchnorm/gamma
(<tf.Tensor 'mixed_7/tower_1/conv_4/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_4/batchnorm/moving_mean
(<tf.Tensor 'mixed_7/tower_1/conv_4/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_4/batchnorm/moving_variance
(<tf.Tensor 'mixed_7/tower_1/conv_4/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_7/tower_1/conv_4/batchnorm
(<tf.Tensor 'mixed_7/tower_1/conv_4/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_4/CheckNumerics
(<tf.Tensor 'mixed_7/tower_1/conv_4/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_4/control_dependency
(<tf.Tensor 'mixed_7/tower_1/conv_4/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_1/conv_4
(<tf.Tensor 'mixed_7/tower_1/conv_4:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_2/pool
(<tf.Tensor 'mixed_7/tower_2/pool:0' shape=(1, 17, 17, 768) dtype=float32>,)
mixed_7/tower_2/conv/conv2d_params
(<tf.Tensor 'mixed_7/tower_2/conv/conv2d_params:0' shape=(1, 1, 768, 192) dtype=float32>,)
mixed_7/tower_2/conv/Conv2D
(<tf.Tensor 'mixed_7/tower_2/conv/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_2/conv/batchnorm/beta
(<tf.Tensor 'mixed_7/tower_2/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_7/tower_2/conv/batchnorm/gamma
(<tf.Tensor 'mixed_7/tower_2/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_7/tower_2/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_7/tower_2/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_7/tower_2/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_7/tower_2/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_7/tower_2/conv/batchnorm
(<tf.Tensor 'mixed_7/tower_2/conv/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_2/conv/CheckNumerics
(<tf.Tensor 'mixed_7/tower_2/conv/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_2/conv/control_dependency
(<tf.Tensor 'mixed_7/tower_2/conv/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/tower_2/conv
(<tf.Tensor 'mixed_7/tower_2/conv:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_7/join/concat_dim
(<tf.Tensor 'mixed_7/join/concat_dim:0' shape=() dtype=int32>,)
mixed_7/join
(<tf.Tensor 'mixed_7/join:0' shape=(1, 17, 17, 768) dtype=float32>,)
mixed_8/tower/conv/conv2d_params
(<tf.Tensor 'mixed_8/tower/conv/conv2d_params:0' shape=(1, 1, 768, 192) dtype=float32>,)
mixed_8/tower/conv/Conv2D
(<tf.Tensor 'mixed_8/tower/conv/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower/conv/batchnorm/beta
(<tf.Tensor 'mixed_8/tower/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_8/tower/conv/batchnorm/gamma
(<tf.Tensor 'mixed_8/tower/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_8/tower/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_8/tower/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_8/tower/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_8/tower/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_8/tower/conv/batchnorm
(<tf.Tensor 'mixed_8/tower/conv/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower/conv/CheckNumerics
(<tf.Tensor 'mixed_8/tower/conv/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower/conv/control_dependency
(<tf.Tensor 'mixed_8/tower/conv/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower/conv
(<tf.Tensor 'mixed_8/tower/conv:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower/conv_1/conv2d_params
(<tf.Tensor 'mixed_8/tower/conv_1/conv2d_params:0' shape=(3, 3, 192, 320) dtype=float32>,)
mixed_8/tower/conv_1/Conv2D
(<tf.Tensor 'mixed_8/tower/conv_1/Conv2D:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_8/tower/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_8/tower/conv_1/batchnorm/beta:0' shape=(320,) dtype=float32>,)
mixed_8/tower/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_8/tower/conv_1/batchnorm/gamma:0' shape=(320,) dtype=float32>,)
mixed_8/tower/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_8/tower/conv_1/batchnorm/moving_mean:0' shape=(320,) dtype=float32>,)
mixed_8/tower/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_8/tower/conv_1/batchnorm/moving_variance:0' shape=(320,) dtype=float32>,)
mixed_8/tower/conv_1/batchnorm
(<tf.Tensor 'mixed_8/tower/conv_1/batchnorm:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_8/tower/conv_1/CheckNumerics
(<tf.Tensor 'mixed_8/tower/conv_1/CheckNumerics:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_8/tower/conv_1/control_dependency
(<tf.Tensor 'mixed_8/tower/conv_1/control_dependency:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_8/tower/conv_1
(<tf.Tensor 'mixed_8/tower/conv_1:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_8/tower_1/conv/conv2d_params
(<tf.Tensor 'mixed_8/tower_1/conv/conv2d_params:0' shape=(1, 1, 768, 192) dtype=float32>,)
mixed_8/tower_1/conv/Conv2D
(<tf.Tensor 'mixed_8/tower_1/conv/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv/batchnorm/beta
(<tf.Tensor 'mixed_8/tower_1/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv/batchnorm/gamma
(<tf.Tensor 'mixed_8/tower_1/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_8/tower_1/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_8/tower_1/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv/batchnorm
(<tf.Tensor 'mixed_8/tower_1/conv/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv/CheckNumerics
(<tf.Tensor 'mixed_8/tower_1/conv/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv/control_dependency
(<tf.Tensor 'mixed_8/tower_1/conv/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv
(<tf.Tensor 'mixed_8/tower_1/conv:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv_1/conv2d_params
(<tf.Tensor 'mixed_8/tower_1/conv_1/conv2d_params:0' shape=(1, 7, 192, 192) dtype=float32>,)
mixed_8/tower_1/conv_1/Conv2D
(<tf.Tensor 'mixed_8/tower_1/conv_1/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_8/tower_1/conv_1/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_8/tower_1/conv_1/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_8/tower_1/conv_1/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_8/tower_1/conv_1/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv_1/batchnorm
(<tf.Tensor 'mixed_8/tower_1/conv_1/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv_1/CheckNumerics
(<tf.Tensor 'mixed_8/tower_1/conv_1/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv_1/control_dependency
(<tf.Tensor 'mixed_8/tower_1/conv_1/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv_1
(<tf.Tensor 'mixed_8/tower_1/conv_1:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv_2/conv2d_params
(<tf.Tensor 'mixed_8/tower_1/conv_2/conv2d_params:0' shape=(7, 1, 192, 192) dtype=float32>,)
mixed_8/tower_1/conv_2/Conv2D
(<tf.Tensor 'mixed_8/tower_1/conv_2/Conv2D:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv_2/batchnorm/beta
(<tf.Tensor 'mixed_8/tower_1/conv_2/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv_2/batchnorm/gamma
(<tf.Tensor 'mixed_8/tower_1/conv_2/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv_2/batchnorm/moving_mean
(<tf.Tensor 'mixed_8/tower_1/conv_2/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv_2/batchnorm/moving_variance
(<tf.Tensor 'mixed_8/tower_1/conv_2/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv_2/batchnorm
(<tf.Tensor 'mixed_8/tower_1/conv_2/batchnorm:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv_2/CheckNumerics
(<tf.Tensor 'mixed_8/tower_1/conv_2/CheckNumerics:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv_2/control_dependency
(<tf.Tensor 'mixed_8/tower_1/conv_2/control_dependency:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv_2
(<tf.Tensor 'mixed_8/tower_1/conv_2:0' shape=(1, 17, 17, 192) dtype=float32>,)
mixed_8/tower_1/conv_3/conv2d_params
(<tf.Tensor 'mixed_8/tower_1/conv_3/conv2d_params:0' shape=(3, 3, 192, 192) dtype=float32>,)
mixed_8/tower_1/conv_3/Conv2D
(<tf.Tensor 'mixed_8/tower_1/conv_3/Conv2D:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_8/tower_1/conv_3/batchnorm/beta
(<tf.Tensor 'mixed_8/tower_1/conv_3/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv_3/batchnorm/gamma
(<tf.Tensor 'mixed_8/tower_1/conv_3/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv_3/batchnorm/moving_mean
(<tf.Tensor 'mixed_8/tower_1/conv_3/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv_3/batchnorm/moving_variance
(<tf.Tensor 'mixed_8/tower_1/conv_3/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_8/tower_1/conv_3/batchnorm
(<tf.Tensor 'mixed_8/tower_1/conv_3/batchnorm:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_8/tower_1/conv_3/CheckNumerics
(<tf.Tensor 'mixed_8/tower_1/conv_3/CheckNumerics:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_8/tower_1/conv_3/control_dependency
(<tf.Tensor 'mixed_8/tower_1/conv_3/control_dependency:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_8/tower_1/conv_3
(<tf.Tensor 'mixed_8/tower_1/conv_3:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_8/pool/CheckNumerics
(<tf.Tensor 'mixed_8/pool/CheckNumerics:0' shape=(1, 17, 17, 768) dtype=float32>,)
mixed_8/pool/control_dependency
(<tf.Tensor 'mixed_8/pool/control_dependency:0' shape=(1, 17, 17, 768) dtype=float32>,)
mixed_8/pool
(<tf.Tensor 'mixed_8/pool:0' shape=(1, 8, 8, 768) dtype=float32>,)
mixed_8/join/concat_dim
(<tf.Tensor 'mixed_8/join/concat_dim:0' shape=() dtype=int32>,)
mixed_8/join
(<tf.Tensor 'mixed_8/join:0' shape=(1, 8, 8, 1280) dtype=float32>,)
mixed_9/conv/conv2d_params
(<tf.Tensor 'mixed_9/conv/conv2d_params:0' shape=(1, 1, 1280, 320) dtype=float32>,)
mixed_9/conv/Conv2D
(<tf.Tensor 'mixed_9/conv/Conv2D:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_9/conv/batchnorm/beta
(<tf.Tensor 'mixed_9/conv/batchnorm/beta:0' shape=(320,) dtype=float32>,)
mixed_9/conv/batchnorm/gamma
(<tf.Tensor 'mixed_9/conv/batchnorm/gamma:0' shape=(320,) dtype=float32>,)
mixed_9/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_9/conv/batchnorm/moving_mean:0' shape=(320,) dtype=float32>,)
mixed_9/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_9/conv/batchnorm/moving_variance:0' shape=(320,) dtype=float32>,)
mixed_9/conv/batchnorm
(<tf.Tensor 'mixed_9/conv/batchnorm:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_9/conv/CheckNumerics
(<tf.Tensor 'mixed_9/conv/CheckNumerics:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_9/conv/control_dependency
(<tf.Tensor 'mixed_9/conv/control_dependency:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_9/conv
(<tf.Tensor 'mixed_9/conv:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_9/tower/conv/conv2d_params
(<tf.Tensor 'mixed_9/tower/conv/conv2d_params:0' shape=(1, 1, 1280, 384) dtype=float32>,)
mixed_9/tower/conv/Conv2D
(<tf.Tensor 'mixed_9/tower/conv/Conv2D:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/conv/batchnorm/beta
(<tf.Tensor 'mixed_9/tower/conv/batchnorm/beta:0' shape=(384,) dtype=float32>,)
mixed_9/tower/conv/batchnorm/gamma
(<tf.Tensor 'mixed_9/tower/conv/batchnorm/gamma:0' shape=(384,) dtype=float32>,)
mixed_9/tower/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_9/tower/conv/batchnorm/moving_mean:0' shape=(384,) dtype=float32>,)
mixed_9/tower/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_9/tower/conv/batchnorm/moving_variance:0' shape=(384,) dtype=float32>,)
mixed_9/tower/conv/batchnorm
(<tf.Tensor 'mixed_9/tower/conv/batchnorm:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/conv/CheckNumerics
(<tf.Tensor 'mixed_9/tower/conv/CheckNumerics:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/conv/control_dependency
(<tf.Tensor 'mixed_9/tower/conv/control_dependency:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/conv
(<tf.Tensor 'mixed_9/tower/conv:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/mixed/conv/conv2d_params
(<tf.Tensor 'mixed_9/tower/mixed/conv/conv2d_params:0' shape=(1, 3, 384, 384) dtype=float32>,)
mixed_9/tower/mixed/conv/Conv2D
(<tf.Tensor 'mixed_9/tower/mixed/conv/Conv2D:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/mixed/conv/batchnorm/beta
(<tf.Tensor 'mixed_9/tower/mixed/conv/batchnorm/beta:0' shape=(384,) dtype=float32>,)
mixed_9/tower/mixed/conv/batchnorm/gamma
(<tf.Tensor 'mixed_9/tower/mixed/conv/batchnorm/gamma:0' shape=(384,) dtype=float32>,)
mixed_9/tower/mixed/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_9/tower/mixed/conv/batchnorm/moving_mean:0' shape=(384,) dtype=float32>,)
mixed_9/tower/mixed/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_9/tower/mixed/conv/batchnorm/moving_variance:0' shape=(384,) dtype=float32>,)
mixed_9/tower/mixed/conv/batchnorm
(<tf.Tensor 'mixed_9/tower/mixed/conv/batchnorm:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/mixed/conv/CheckNumerics
(<tf.Tensor 'mixed_9/tower/mixed/conv/CheckNumerics:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/mixed/conv/control_dependency
(<tf.Tensor 'mixed_9/tower/mixed/conv/control_dependency:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/mixed/conv
(<tf.Tensor 'mixed_9/tower/mixed/conv:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/mixed/conv_1/conv2d_params
(<tf.Tensor 'mixed_9/tower/mixed/conv_1/conv2d_params:0' shape=(3, 1, 384, 384) dtype=float32>,)
mixed_9/tower/mixed/conv_1/Conv2D
(<tf.Tensor 'mixed_9/tower/mixed/conv_1/Conv2D:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/mixed/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_9/tower/mixed/conv_1/batchnorm/beta:0' shape=(384,) dtype=float32>,)
mixed_9/tower/mixed/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_9/tower/mixed/conv_1/batchnorm/gamma:0' shape=(384,) dtype=float32>,)
mixed_9/tower/mixed/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_9/tower/mixed/conv_1/batchnorm/moving_mean:0' shape=(384,) dtype=float32>,)
mixed_9/tower/mixed/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_9/tower/mixed/conv_1/batchnorm/moving_variance:0' shape=(384,) dtype=float32>,)
mixed_9/tower/mixed/conv_1/batchnorm
(<tf.Tensor 'mixed_9/tower/mixed/conv_1/batchnorm:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/mixed/conv_1/CheckNumerics
(<tf.Tensor 'mixed_9/tower/mixed/conv_1/CheckNumerics:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/mixed/conv_1/control_dependency
(<tf.Tensor 'mixed_9/tower/mixed/conv_1/control_dependency:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower/mixed/conv_1
(<tf.Tensor 'mixed_9/tower/mixed/conv_1:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/conv/conv2d_params
(<tf.Tensor 'mixed_9/tower_1/conv/conv2d_params:0' shape=(1, 1, 1280, 448) dtype=float32>,)
mixed_9/tower_1/conv/Conv2D
(<tf.Tensor 'mixed_9/tower_1/conv/Conv2D:0' shape=(1, 8, 8, 448) dtype=float32>,)
mixed_9/tower_1/conv/batchnorm/beta
(<tf.Tensor 'mixed_9/tower_1/conv/batchnorm/beta:0' shape=(448,) dtype=float32>,)
mixed_9/tower_1/conv/batchnorm/gamma
(<tf.Tensor 'mixed_9/tower_1/conv/batchnorm/gamma:0' shape=(448,) dtype=float32>,)
mixed_9/tower_1/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_9/tower_1/conv/batchnorm/moving_mean:0' shape=(448,) dtype=float32>,)
mixed_9/tower_1/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_9/tower_1/conv/batchnorm/moving_variance:0' shape=(448,) dtype=float32>,)
mixed_9/tower_1/conv/batchnorm
(<tf.Tensor 'mixed_9/tower_1/conv/batchnorm:0' shape=(1, 8, 8, 448) dtype=float32>,)
mixed_9/tower_1/conv/CheckNumerics
(<tf.Tensor 'mixed_9/tower_1/conv/CheckNumerics:0' shape=(1, 8, 8, 448) dtype=float32>,)
mixed_9/tower_1/conv/control_dependency
(<tf.Tensor 'mixed_9/tower_1/conv/control_dependency:0' shape=(1, 8, 8, 448) dtype=float32>,)
mixed_9/tower_1/conv
(<tf.Tensor 'mixed_9/tower_1/conv:0' shape=(1, 8, 8, 448) dtype=float32>,)
mixed_9/tower_1/conv_1/conv2d_params
(<tf.Tensor 'mixed_9/tower_1/conv_1/conv2d_params:0' shape=(3, 3, 448, 384) dtype=float32>,)
mixed_9/tower_1/conv_1/Conv2D
(<tf.Tensor 'mixed_9/tower_1/conv_1/Conv2D:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_9/tower_1/conv_1/batchnorm/beta:0' shape=(384,) dtype=float32>,)
mixed_9/tower_1/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_9/tower_1/conv_1/batchnorm/gamma:0' shape=(384,) dtype=float32>,)
mixed_9/tower_1/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_9/tower_1/conv_1/batchnorm/moving_mean:0' shape=(384,) dtype=float32>,)
mixed_9/tower_1/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_9/tower_1/conv_1/batchnorm/moving_variance:0' shape=(384,) dtype=float32>,)
mixed_9/tower_1/conv_1/batchnorm
(<tf.Tensor 'mixed_9/tower_1/conv_1/batchnorm:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/conv_1/CheckNumerics
(<tf.Tensor 'mixed_9/tower_1/conv_1/CheckNumerics:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/conv_1/control_dependency
(<tf.Tensor 'mixed_9/tower_1/conv_1/control_dependency:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/conv_1
(<tf.Tensor 'mixed_9/tower_1/conv_1:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/mixed/conv/conv2d_params
(<tf.Tensor 'mixed_9/tower_1/mixed/conv/conv2d_params:0' shape=(1, 3, 384, 384) dtype=float32>,)
mixed_9/tower_1/mixed/conv/Conv2D
(<tf.Tensor 'mixed_9/tower_1/mixed/conv/Conv2D:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/mixed/conv/batchnorm/beta
(<tf.Tensor 'mixed_9/tower_1/mixed/conv/batchnorm/beta:0' shape=(384,) dtype=float32>,)
mixed_9/tower_1/mixed/conv/batchnorm/gamma
(<tf.Tensor 'mixed_9/tower_1/mixed/conv/batchnorm/gamma:0' shape=(384,) dtype=float32>,)
mixed_9/tower_1/mixed/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_9/tower_1/mixed/conv/batchnorm/moving_mean:0' shape=(384,) dtype=float32>,)
mixed_9/tower_1/mixed/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_9/tower_1/mixed/conv/batchnorm/moving_variance:0' shape=(384,) dtype=float32>,)
mixed_9/tower_1/mixed/conv/batchnorm
(<tf.Tensor 'mixed_9/tower_1/mixed/conv/batchnorm:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/mixed/conv/CheckNumerics
(<tf.Tensor 'mixed_9/tower_1/mixed/conv/CheckNumerics:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/mixed/conv/control_dependency
(<tf.Tensor 'mixed_9/tower_1/mixed/conv/control_dependency:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/mixed/conv
(<tf.Tensor 'mixed_9/tower_1/mixed/conv:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/mixed/conv_1/conv2d_params
(<tf.Tensor 'mixed_9/tower_1/mixed/conv_1/conv2d_params:0' shape=(3, 1, 384, 384) dtype=float32>,)
mixed_9/tower_1/mixed/conv_1/Conv2D
(<tf.Tensor 'mixed_9/tower_1/mixed/conv_1/Conv2D:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/mixed/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_9/tower_1/mixed/conv_1/batchnorm/beta:0' shape=(384,) dtype=float32>,)
mixed_9/tower_1/mixed/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_9/tower_1/mixed/conv_1/batchnorm/gamma:0' shape=(384,) dtype=float32>,)
mixed_9/tower_1/mixed/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_9/tower_1/mixed/conv_1/batchnorm/moving_mean:0' shape=(384,) dtype=float32>,)
mixed_9/tower_1/mixed/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_9/tower_1/mixed/conv_1/batchnorm/moving_variance:0' shape=(384,) dtype=float32>,)
mixed_9/tower_1/mixed/conv_1/batchnorm
(<tf.Tensor 'mixed_9/tower_1/mixed/conv_1/batchnorm:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/mixed/conv_1/CheckNumerics
(<tf.Tensor 'mixed_9/tower_1/mixed/conv_1/CheckNumerics:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/mixed/conv_1/control_dependency
(<tf.Tensor 'mixed_9/tower_1/mixed/conv_1/control_dependency:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_1/mixed/conv_1
(<tf.Tensor 'mixed_9/tower_1/mixed/conv_1:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_9/tower_2/pool
(<tf.Tensor 'mixed_9/tower_2/pool:0' shape=(1, 8, 8, 1280) dtype=float32>,)
mixed_9/tower_2/conv/conv2d_params
(<tf.Tensor 'mixed_9/tower_2/conv/conv2d_params:0' shape=(1, 1, 1280, 192) dtype=float32>,)
mixed_9/tower_2/conv/Conv2D
(<tf.Tensor 'mixed_9/tower_2/conv/Conv2D:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_9/tower_2/conv/batchnorm/beta
(<tf.Tensor 'mixed_9/tower_2/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_9/tower_2/conv/batchnorm/gamma
(<tf.Tensor 'mixed_9/tower_2/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_9/tower_2/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_9/tower_2/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_9/tower_2/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_9/tower_2/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_9/tower_2/conv/batchnorm
(<tf.Tensor 'mixed_9/tower_2/conv/batchnorm:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_9/tower_2/conv/CheckNumerics
(<tf.Tensor 'mixed_9/tower_2/conv/CheckNumerics:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_9/tower_2/conv/control_dependency
(<tf.Tensor 'mixed_9/tower_2/conv/control_dependency:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_9/tower_2/conv
(<tf.Tensor 'mixed_9/tower_2/conv:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_9/join/concat_dim
(<tf.Tensor 'mixed_9/join/concat_dim:0' shape=() dtype=int32>,)
mixed_9/join
(<tf.Tensor 'mixed_9/join:0' shape=(1, 8, 8, 2048) dtype=float32>,)
mixed_10/conv/conv2d_params
(<tf.Tensor 'mixed_10/conv/conv2d_params:0' shape=(1, 1, 2048, 320) dtype=float32>,)
mixed_10/conv/Conv2D
(<tf.Tensor 'mixed_10/conv/Conv2D:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_10/conv/batchnorm/beta
(<tf.Tensor 'mixed_10/conv/batchnorm/beta:0' shape=(320,) dtype=float32>,)
mixed_10/conv/batchnorm/gamma
(<tf.Tensor 'mixed_10/conv/batchnorm/gamma:0' shape=(320,) dtype=float32>,)
mixed_10/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_10/conv/batchnorm/moving_mean:0' shape=(320,) dtype=float32>,)
mixed_10/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_10/conv/batchnorm/moving_variance:0' shape=(320,) dtype=float32>,)
mixed_10/conv/batchnorm
(<tf.Tensor 'mixed_10/conv/batchnorm:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_10/conv/CheckNumerics
(<tf.Tensor 'mixed_10/conv/CheckNumerics:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_10/conv/control_dependency
(<tf.Tensor 'mixed_10/conv/control_dependency:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_10/conv
(<tf.Tensor 'mixed_10/conv:0' shape=(1, 8, 8, 320) dtype=float32>,)
mixed_10/tower/conv/conv2d_params
(<tf.Tensor 'mixed_10/tower/conv/conv2d_params:0' shape=(1, 1, 2048, 384) dtype=float32>,)
mixed_10/tower/conv/Conv2D
(<tf.Tensor 'mixed_10/tower/conv/Conv2D:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/conv/batchnorm/beta
(<tf.Tensor 'mixed_10/tower/conv/batchnorm/beta:0' shape=(384,) dtype=float32>,)
mixed_10/tower/conv/batchnorm/gamma
(<tf.Tensor 'mixed_10/tower/conv/batchnorm/gamma:0' shape=(384,) dtype=float32>,)
mixed_10/tower/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_10/tower/conv/batchnorm/moving_mean:0' shape=(384,) dtype=float32>,)
mixed_10/tower/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_10/tower/conv/batchnorm/moving_variance:0' shape=(384,) dtype=float32>,)
mixed_10/tower/conv/batchnorm
(<tf.Tensor 'mixed_10/tower/conv/batchnorm:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/conv/CheckNumerics
(<tf.Tensor 'mixed_10/tower/conv/CheckNumerics:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/conv/control_dependency
(<tf.Tensor 'mixed_10/tower/conv/control_dependency:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/conv
(<tf.Tensor 'mixed_10/tower/conv:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/mixed/conv/conv2d_params
(<tf.Tensor 'mixed_10/tower/mixed/conv/conv2d_params:0' shape=(1, 3, 384, 384) dtype=float32>,)
mixed_10/tower/mixed/conv/Conv2D
(<tf.Tensor 'mixed_10/tower/mixed/conv/Conv2D:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/mixed/conv/batchnorm/beta
(<tf.Tensor 'mixed_10/tower/mixed/conv/batchnorm/beta:0' shape=(384,) dtype=float32>,)
mixed_10/tower/mixed/conv/batchnorm/gamma
(<tf.Tensor 'mixed_10/tower/mixed/conv/batchnorm/gamma:0' shape=(384,) dtype=float32>,)
mixed_10/tower/mixed/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_10/tower/mixed/conv/batchnorm/moving_mean:0' shape=(384,) dtype=float32>,)
mixed_10/tower/mixed/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_10/tower/mixed/conv/batchnorm/moving_variance:0' shape=(384,) dtype=float32>,)
mixed_10/tower/mixed/conv/batchnorm
(<tf.Tensor 'mixed_10/tower/mixed/conv/batchnorm:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/mixed/conv/CheckNumerics
(<tf.Tensor 'mixed_10/tower/mixed/conv/CheckNumerics:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/mixed/conv/control_dependency
(<tf.Tensor 'mixed_10/tower/mixed/conv/control_dependency:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/mixed/conv
(<tf.Tensor 'mixed_10/tower/mixed/conv:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/mixed/conv_1/conv2d_params
(<tf.Tensor 'mixed_10/tower/mixed/conv_1/conv2d_params:0' shape=(3, 1, 384, 384) dtype=float32>,)
mixed_10/tower/mixed/conv_1/Conv2D
(<tf.Tensor 'mixed_10/tower/mixed/conv_1/Conv2D:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/mixed/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_10/tower/mixed/conv_1/batchnorm/beta:0' shape=(384,) dtype=float32>,)
mixed_10/tower/mixed/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_10/tower/mixed/conv_1/batchnorm/gamma:0' shape=(384,) dtype=float32>,)
mixed_10/tower/mixed/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_10/tower/mixed/conv_1/batchnorm/moving_mean:0' shape=(384,) dtype=float32>,)
mixed_10/tower/mixed/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_10/tower/mixed/conv_1/batchnorm/moving_variance:0' shape=(384,) dtype=float32>,)
mixed_10/tower/mixed/conv_1/batchnorm
(<tf.Tensor 'mixed_10/tower/mixed/conv_1/batchnorm:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/mixed/conv_1/CheckNumerics
(<tf.Tensor 'mixed_10/tower/mixed/conv_1/CheckNumerics:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/mixed/conv_1/control_dependency
(<tf.Tensor 'mixed_10/tower/mixed/conv_1/control_dependency:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower/mixed/conv_1
(<tf.Tensor 'mixed_10/tower/mixed/conv_1:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/conv/conv2d_params
(<tf.Tensor 'mixed_10/tower_1/conv/conv2d_params:0' shape=(1, 1, 2048, 448) dtype=float32>,)
mixed_10/tower_1/conv/Conv2D
(<tf.Tensor 'mixed_10/tower_1/conv/Conv2D:0' shape=(1, 8, 8, 448) dtype=float32>,)
mixed_10/tower_1/conv/batchnorm/beta
(<tf.Tensor 'mixed_10/tower_1/conv/batchnorm/beta:0' shape=(448,) dtype=float32>,)
mixed_10/tower_1/conv/batchnorm/gamma
(<tf.Tensor 'mixed_10/tower_1/conv/batchnorm/gamma:0' shape=(448,) dtype=float32>,)
mixed_10/tower_1/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_10/tower_1/conv/batchnorm/moving_mean:0' shape=(448,) dtype=float32>,)
mixed_10/tower_1/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_10/tower_1/conv/batchnorm/moving_variance:0' shape=(448,) dtype=float32>,)
mixed_10/tower_1/conv/batchnorm
(<tf.Tensor 'mixed_10/tower_1/conv/batchnorm:0' shape=(1, 8, 8, 448) dtype=float32>,)
mixed_10/tower_1/conv/CheckNumerics
(<tf.Tensor 'mixed_10/tower_1/conv/CheckNumerics:0' shape=(1, 8, 8, 448) dtype=float32>,)
mixed_10/tower_1/conv/control_dependency
(<tf.Tensor 'mixed_10/tower_1/conv/control_dependency:0' shape=(1, 8, 8, 448) dtype=float32>,)
mixed_10/tower_1/conv
(<tf.Tensor 'mixed_10/tower_1/conv:0' shape=(1, 8, 8, 448) dtype=float32>,)
mixed_10/tower_1/conv_1/conv2d_params
(<tf.Tensor 'mixed_10/tower_1/conv_1/conv2d_params:0' shape=(3, 3, 448, 384) dtype=float32>,)
mixed_10/tower_1/conv_1/Conv2D
(<tf.Tensor 'mixed_10/tower_1/conv_1/Conv2D:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_10/tower_1/conv_1/batchnorm/beta:0' shape=(384,) dtype=float32>,)
mixed_10/tower_1/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_10/tower_1/conv_1/batchnorm/gamma:0' shape=(384,) dtype=float32>,)
mixed_10/tower_1/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_10/tower_1/conv_1/batchnorm/moving_mean:0' shape=(384,) dtype=float32>,)
mixed_10/tower_1/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_10/tower_1/conv_1/batchnorm/moving_variance:0' shape=(384,) dtype=float32>,)
mixed_10/tower_1/conv_1/batchnorm
(<tf.Tensor 'mixed_10/tower_1/conv_1/batchnorm:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/conv_1/CheckNumerics
(<tf.Tensor 'mixed_10/tower_1/conv_1/CheckNumerics:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/conv_1/control_dependency
(<tf.Tensor 'mixed_10/tower_1/conv_1/control_dependency:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/conv_1
(<tf.Tensor 'mixed_10/tower_1/conv_1:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/mixed/conv/conv2d_params
(<tf.Tensor 'mixed_10/tower_1/mixed/conv/conv2d_params:0' shape=(1, 3, 384, 384) dtype=float32>,)
mixed_10/tower_1/mixed/conv/Conv2D
(<tf.Tensor 'mixed_10/tower_1/mixed/conv/Conv2D:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/mixed/conv/batchnorm/beta
(<tf.Tensor 'mixed_10/tower_1/mixed/conv/batchnorm/beta:0' shape=(384,) dtype=float32>,)
mixed_10/tower_1/mixed/conv/batchnorm/gamma
(<tf.Tensor 'mixed_10/tower_1/mixed/conv/batchnorm/gamma:0' shape=(384,) dtype=float32>,)
mixed_10/tower_1/mixed/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_10/tower_1/mixed/conv/batchnorm/moving_mean:0' shape=(384,) dtype=float32>,)
mixed_10/tower_1/mixed/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_10/tower_1/mixed/conv/batchnorm/moving_variance:0' shape=(384,) dtype=float32>,)
mixed_10/tower_1/mixed/conv/batchnorm
(<tf.Tensor 'mixed_10/tower_1/mixed/conv/batchnorm:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/mixed/conv/CheckNumerics
(<tf.Tensor 'mixed_10/tower_1/mixed/conv/CheckNumerics:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/mixed/conv/control_dependency
(<tf.Tensor 'mixed_10/tower_1/mixed/conv/control_dependency:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/mixed/conv
(<tf.Tensor 'mixed_10/tower_1/mixed/conv:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/mixed/conv_1/conv2d_params
(<tf.Tensor 'mixed_10/tower_1/mixed/conv_1/conv2d_params:0' shape=(3, 1, 384, 384) dtype=float32>,)
mixed_10/tower_1/mixed/conv_1/Conv2D
(<tf.Tensor 'mixed_10/tower_1/mixed/conv_1/Conv2D:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/mixed/conv_1/batchnorm/beta
(<tf.Tensor 'mixed_10/tower_1/mixed/conv_1/batchnorm/beta:0' shape=(384,) dtype=float32>,)
mixed_10/tower_1/mixed/conv_1/batchnorm/gamma
(<tf.Tensor 'mixed_10/tower_1/mixed/conv_1/batchnorm/gamma:0' shape=(384,) dtype=float32>,)
mixed_10/tower_1/mixed/conv_1/batchnorm/moving_mean
(<tf.Tensor 'mixed_10/tower_1/mixed/conv_1/batchnorm/moving_mean:0' shape=(384,) dtype=float32>,)
mixed_10/tower_1/mixed/conv_1/batchnorm/moving_variance
(<tf.Tensor 'mixed_10/tower_1/mixed/conv_1/batchnorm/moving_variance:0' shape=(384,) dtype=float32>,)
mixed_10/tower_1/mixed/conv_1/batchnorm
(<tf.Tensor 'mixed_10/tower_1/mixed/conv_1/batchnorm:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/mixed/conv_1/CheckNumerics
(<tf.Tensor 'mixed_10/tower_1/mixed/conv_1/CheckNumerics:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/mixed/conv_1/control_dependency
(<tf.Tensor 'mixed_10/tower_1/mixed/conv_1/control_dependency:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_1/mixed/conv_1
(<tf.Tensor 'mixed_10/tower_1/mixed/conv_1:0' shape=(1, 8, 8, 384) dtype=float32>,)
mixed_10/tower_2/pool/CheckNumerics
(<tf.Tensor 'mixed_10/tower_2/pool/CheckNumerics:0' shape=(1, 8, 8, 2048) dtype=float32>,)
mixed_10/tower_2/pool/control_dependency
(<tf.Tensor 'mixed_10/tower_2/pool/control_dependency:0' shape=(1, 8, 8, 2048) dtype=float32>,)
mixed_10/tower_2/pool
(<tf.Tensor 'mixed_10/tower_2/pool:0' shape=(1, 8, 8, 2048) dtype=float32>,)
mixed_10/tower_2/conv/conv2d_params
(<tf.Tensor 'mixed_10/tower_2/conv/conv2d_params:0' shape=(1, 1, 2048, 192) dtype=float32>,)
mixed_10/tower_2/conv/Conv2D
(<tf.Tensor 'mixed_10/tower_2/conv/Conv2D:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_10/tower_2/conv/batchnorm/beta
(<tf.Tensor 'mixed_10/tower_2/conv/batchnorm/beta:0' shape=(192,) dtype=float32>,)
mixed_10/tower_2/conv/batchnorm/gamma
(<tf.Tensor 'mixed_10/tower_2/conv/batchnorm/gamma:0' shape=(192,) dtype=float32>,)
mixed_10/tower_2/conv/batchnorm/moving_mean
(<tf.Tensor 'mixed_10/tower_2/conv/batchnorm/moving_mean:0' shape=(192,) dtype=float32>,)
mixed_10/tower_2/conv/batchnorm/moving_variance
(<tf.Tensor 'mixed_10/tower_2/conv/batchnorm/moving_variance:0' shape=(192,) dtype=float32>,)
mixed_10/tower_2/conv/batchnorm
(<tf.Tensor 'mixed_10/tower_2/conv/batchnorm:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_10/tower_2/conv/CheckNumerics
(<tf.Tensor 'mixed_10/tower_2/conv/CheckNumerics:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_10/tower_2/conv/control_dependency
(<tf.Tensor 'mixed_10/tower_2/conv/control_dependency:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_10/tower_2/conv
(<tf.Tensor 'mixed_10/tower_2/conv:0' shape=(1, 8, 8, 192) dtype=float32>,)
mixed_10/join/concat_dim
(<tf.Tensor 'mixed_10/join/concat_dim:0' shape=() dtype=int32>,)
mixed_10/join
(<tf.Tensor 'mixed_10/join:0' shape=(1, 8, 8, 2048) dtype=float32>,)
pool_3
(<tf.Tensor 'pool_3:0' shape=(1, 1, 1, 2048) dtype=float32>,)
pool_3/_reshape/shape
(<tf.Tensor 'pool_3/_reshape/shape:0' shape=(2,) dtype=int32>,)
pool_3/_reshape
(<tf.Tensor 'pool_3/_reshape:0' shape=(1, 2048) dtype=float32>,)
softmax/weights
(<tf.Tensor 'softmax/weights:0' shape=(2048, 1008) dtype=float32>,)
softmax/biases
(<tf.Tensor 'softmax/biases:0' shape=(1008,) dtype=float32>,)
softmax/logits/MatMul
(<tf.Tensor 'softmax/logits/MatMul:0' shape=(1, 1008) dtype=float32>,)
softmax/logits
(<tf.Tensor 'softmax/logits:0' shape=(1, 1008) dtype=float32>,)
softmax
(<tf.Tensor 'softmax:0' shape=(1, 1008) dtype=float32>,)

In [9]:

    

#Image data
image = os.path.join(path_model, 'cropped_panda.jpg')
image_data = tf.gfile.FastGFile(image, 'rb').read()


    

In [10]:

    

#Generate predictios from the jpg file
with tf.Session() as sess:
    softmax_tensor = sess.graph.get_tensor_by_name('softmax:0')
    predictions = sess.run(softmax_tensor,
                           {'DecodeJpeg/contents:0': image_data})
predictions = np.squeeze(predictions)

print predictions


    

    




[[  1.00235826e-04   2.51057325e-04   7.98699330e-05 ...,   1.00235244e-04
    1.00236299e-04   1.00236299e-04]]

In [17]:

    

class NodeLookup(object):
  """Converts integer node ID's to human readable labels."""

  def __init__(self,
               label_lookup_path=None,
               uid_lookup_path=None):
    if not label_lookup_path:
      label_lookup_path = os.path.join(
          path_model, 'imagenet_2012_challenge_label_map_proto.pbtxt')
    if not uid_lookup_path:
      uid_lookup_path = os.path.join(
          path_model, 'imagenet_synset_to_human_label_map.txt')
    self.node_lookup = self.load(label_lookup_path, uid_lookup_path)

    
  def load(self, label_lookup_path, uid_lookup_path):
    """Loads a human readable English name for each softmax node.

    Args:
      label_lookup_path: string UID to integer node ID.
      uid_lookup_path: string UID to human-readable string.

    Returns:
      dict from integer node ID to human-readable string.
    """
    if not tf.gfile.Exists(uid_lookup_path):
      tf.logging.fatal('File does not exist %s', uid_lookup_path)
    if not tf.gfile.Exists(label_lookup_path):
      tf.logging.fatal('File does not exist %s', label_lookup_path)

    # Loads mapping from string UID to human-readable string
    proto_as_ascii_lines = tf.gfile.GFile(uid_lookup_path).readlines()
    uid_to_human = {}
    p = re.compile(r'[n\d]*[ \S,]*')
    for line in proto_as_ascii_lines:
      parsed_items = p.findall(line)
      uid = parsed_items[0]
      human_string = parsed_items[2]
      uid_to_human[uid] = human_string

    # Loads mapping from string UID to integer node ID.
    node_id_to_uid = {}
    proto_as_ascii = tf.gfile.GFile(label_lookup_path).readlines()
    for line in proto_as_ascii:
      if line.startswith('  target_class:'):
        target_class = int(line.split(': ')[1])
      if line.startswith('  target_class_string:'):
        target_class_string = line.split(': ')[1]
        node_id_to_uid[target_class] = target_class_string[1:-2]

    # Loads the final mapping of integer node ID to human-readable string
    node_id_to_name = {}
    for key, val in node_id_to_uid.items():
      if val not in uid_to_human:
        tf.logging.fatal('Failed to locate: %s', val)
      name = uid_to_human[val]
      node_id_to_name[key] = name

    return node_id_to_name


  def id_to_string(self, node_id):
    if node_id not in self.node_lookup:
      return ''
    return self.node_lookup[node_id]


    

In [18]:

    

# Creates node ID --> English string lookup.
node_lookup = NodeLookup()

top_k = predictions.argsort()[-5:][::-1]
for node_id in top_k:
    human_string = node_lookup.id_to_string(node_id)
    score = predictions[node_id]
    print('%s (score = %.5f)' % (human_string, score))


    

    




giant panda, panda, panda bear, coon bear, Ailuropoda melanoleuca (score = 0.89233)
indri, indris, Indri indri, Indri brevicaudatus (score = 0.00859)
lesser panda, red panda, panda, bear cat, cat bear, Ailurus fulgens (score = 0.00264)
custard apple (score = 0.00141)
earthstar (score = 0.00107)

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]: