Enregistrer le modèle tensorflow: variables non trouvées

Question

J'essaie d'enregistrer et de restaurer un modèle tensorflow en utilisant tf.train.Saver. Je pense que je l'ai fait la partie de sauvegarde correctement, mais lorsque je tente de restaurer le modèle que je reçois cette erreur:

2017-07-03 15:55:14.824767: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key train/beta2_power not found in checkpoint 
2017-07-03 15:55:14.824796: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases-outputlayer not found in checkpoint 
2017-07-03 15:55:14.825913: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases-outputlayer/Adam not found in checkpoint 
2017-07-03 15:55:14.826588: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases-outputlayer/Adam_1 not found in checkpoint 
2017-07-03 15:55:14.827369: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key train/beta1_power not found in checkpoint 
2017-07-03 15:55:14.828101: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases1 not found in checkpoint 
2017-07-03 15:55:14.828973: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases1/Adam not found in checkpoint 
2017-07-03 15:55:14.829151: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases1/Adam_1 not found in checkpoint 
2017-07-03 15:55:14.830308: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights5/Adam_1 not found in checkpoint 
2017-07-03 15:55:14.830590: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases2 not found in checkpoint 
2017-07-03 15:55:14.831279: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases2/Adam not found in checkpoint 
2017-07-03 15:55:14.832268: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases2/Adam_1 not found in checkpoint 
2017-07-03 15:55:14.832558: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights5/Adam not found in checkpoint 
2017-07-03 15:55:14.833052: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases3 not found in checkpoint 
2017-07-03 15:55:14.834195: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases3/Adam not found in checkpoint 
2017-07-03 15:55:14.834228: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases3/Adam_1 not found in checkpoint 
2017-07-03 15:55:14.834629: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases4 not found in checkpoint 
2017-07-03 15:55:14.835986: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights5 not found in checkpoint 
2017-07-03 15:55:14.836128: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases5 not found in checkpoint 
2017-07-03 15:55:14.836423: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases4/Adam not found in checkpoint 
2017-07-03 15:55:14.837906: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases4/Adam_1 not found in checkpoint 
2017-07-03 15:55:14.838055: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights4/Adam_1 not found in checkpoint 
2017-07-03 15:55:14.838388: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases5/Adam not found in checkpoint 
2017-07-03 15:55:14.839666: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Biases5/Adam_1 not found in checkpoint 
2017-07-03 15:55:14.840299: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights-outputlayer not found in checkpoint 
2017-07-03 15:55:14.840774: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights4/Adam not found in checkpoint 
2017-07-03 15:55:14.841568: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights-outputlayer/Adam not found in checkpoint 
2017-07-03 15:55:14.842312: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights-outputlayer/Adam_1 not found in checkpoint 
2017-07-03 15:55:14.842689: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights1 not found in checkpoint 
2017-07-03 15:55:14.843789: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights4 not found in checkpoint 
2017-07-03 15:55:14.844030: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights1/Adam not found in checkpoint 
2017-07-03 15:55:14.844775: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights1/Adam_1 not found in checkpoint 
2017-07-03 15:55:14.845580: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights2 not found in checkpoint 
2017-07-03 15:55:14.845919: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights3/Adam_1 not found in checkpoint 
2017-07-03 15:55:14.846800: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights2/Adam not found in checkpoint 
2017-07-03 15:55:14.847101: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights2/Adam_1 not found in checkpoint 
2017-07-03 15:55:14.847274: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights3 not found in checkpoint 
2017-07-03 15:55:14.847467: W tensorflow/core/framework/op_kernel.cc:1152] Not found: Key Weights3/Adam not found in checkpoint 
Traceback (most recent call last): 
    File "predict2.py", line 6, in <module> 
    saver.restore(sess,tf.train.latest_checkpoint('./')) 
    File "/anaconda/lib/python2.7/site-packages/tensorflow/python/training/saver.py", line 1457, in restore 
    {self.saver_def.filename_tensor_name: save_path}) 
    File "/anaconda/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 778, in run 
    run_metadata_ptr) 
    File "/anaconda/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 982, in _run 
    feed_dict_string, options, run_metadata) 
    File "/anaconda/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1032, in _do_run 
    target_list, options, run_metadata) 
    File "/anaconda/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1052, in _do_call 
    raise type(e)(node_def, op, message) 
tensorflow.python.framework.errors_impl.NotFoundError: Key Biases-outputlayer not found in checkpoint 
    [[Node: save/RestoreV2 = RestoreV2[dtypes=[DT_FLOAT], _device="/job:localhost/replica:0/task:0/cpu:0"](_recv_save/Const_0, save/RestoreV2/tensor_names, save/RestoreV2/shape_and_slices)]] 

Caused by op u'save/RestoreV2', defined at: 
    File "predict2.py", line 5, in <module> 
    saver = tf.train.import_meta_graph(os.getcwd()+'/models/baseDNN.meta') 
    File "/anaconda/lib/python2.7/site-packages/tensorflow/python/training/saver.py", line 1595, in import_meta_graph 
    **kwargs) 
    File "/anaconda/lib/python2.7/site-packages/tensorflow/python/framework/meta_graph.py", line 499, in import_scoped_meta_graph 
    producer_op_list=producer_op_list) 
    File "/anaconda/lib/python2.7/site-packages/tensorflow/python/framework/importer.py", line 308, in import_graph_def 
    op_def=op_def) 
    File "/anaconda/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2336, in create_op 
    original_op=self._default_original_op, op_def=op_def) 
    File "/anaconda/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1228, in __init__ 
    self._traceback = _extract_stack() 

NotFoundError (see above for traceback): Key Biases-outputlayer not found in checkpoint 
    [[Node: save/RestoreV2 = RestoreV2[dtypes=[DT_FLOAT], _device="/job:localhost/replica:0/task:0/cpu:0"](_recv_save/Const_0, save/RestoreV2/tensor_names, save/RestoreV2/shape_and_slices)]] 

J'ai regardé d'autres questions StackOverflow où ils utilisent import_meta_graph, get_graph_by_tensor et restore charger retour au modèle tensorflow, et j'ai essayé d'adapter cela à mon code, mais je continue d'obtenir ces erreurs où il est dit qu'il ne pouvait pas trouver de variables dans le point de contrôle.

Voici le code pour enregistrer le modèle:

TRAIN_KEEP_PROB = 1.0 
TEST_KEEP_PROB = 1.0 
learning_rate = 0.0001 
ne = 10 

train = 100 
test = 1 
num_nodes = 250 
len_puzzle = 80 
n_nodes_hl1 = num_nodes # hidden layer 1 
n_nodes_hl2 = num_nodes 
n_nodes_hl3 = num_nodes 
n_nodes_hl4 = num_nodes 
n_nodes_hl5 = num_nodes 

n_classes = 4 
batch_size = 100 # load 100 features at a time 

x = tf.placeholder('float',[None,TF_SHAPE],name="x_placeholder") 
y = tf.placeholder('float',name='y_placeholder') 
keep_prob = tf.placeholder('float',name='keep_prob_placeholder') 

def neuralNet(data): 
    hl_1 = {'weights':tf.Variable(tf.random_normal([TF_SHAPE, n_nodes_hl1]),name='Weights1'), 
      'biases':tf.Variable(tf.random_normal([n_nodes_hl1]),name='Biases1')} 

    hl_2 = {'weights':tf.Variable(tf.random_normal([n_nodes_hl1, n_nodes_hl2]),name='Weights2'), 
      'biases':tf.Variable(tf.random_normal([n_nodes_hl2]),name='Biases2')} 

    hl_3 = {'weights':tf.Variable(tf.random_normal([n_nodes_hl2, n_nodes_hl3]),name='Weights3'), 
      'biases':tf.Variable(tf.random_normal([n_nodes_hl3]),name='Biases3')} 

    hl_4 = {'weights':tf.Variable(tf.random_normal([n_nodes_hl3, n_nodes_hl4]),name='Weights4'), 
      'biases':tf.Variable(tf.random_normal([n_nodes_hl4]),name='Biases4')} 

    hl_5 = {'weights':tf.Variable(tf.random_normal([n_nodes_hl4, n_nodes_hl5]),name='Weights5'), 
      'biases':tf.Variable(tf.random_normal([n_nodes_hl5]),name='Biases5')} 

    output_layer = {'weights':tf.Variable(tf.random_normal([n_nodes_hl5, n_classes]),name='Weights-outputlayer'), 
      'biases':tf.Variable(tf.random_normal([n_classes]),name='Biases-outputlayer')} 

    l1 = tf.add(tf.matmul(data, hl_1['weights']), hl_1['biases']) 
    l1 = tf.nn.sigmoid(l1,name='op1') 

    l2 = tf.add(tf.matmul(l1, hl_2['weights']), hl_2['biases']) 
    l2 = tf.nn.sigmoid(l2,name='op2') 

    l3 = tf.add(tf.matmul(l2, hl_3['weights']), hl_3['biases']) 
    l3 = tf.nn.sigmoid(l3,name='op3') 

    l4 = tf.add(tf.matmul(l3, hl_4['weights']), hl_4['biases']) 
    l4 = tf.nn.sigmoid(l4,name='op4') 

    l5 = tf.add(tf.matmul(l4, hl_5['weights']), hl_5['biases']) 
    l5 = tf.nn.sigmoid(l5,name='op5') 

    dropout = tf.nn.dropout(l5,keep_prob, name='op6') 
    ol = tf.add(tf.matmul(dropout, output_layer['weights']), output_layer['biases'], name='op7') 

    return ol 


def train(x): 
    prediction = neuralNet(x) 
    #print prediction 
    with tf.name_scope('cross_entropy'): 
     cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=prediction,labels=y)) 
     tf.summary.scalar('cross_entropy',cost) 

    with tf.name_scope('train'): 
     optimizer = tf.train.AdamOptimizer(learning_rate).minimize(cost) # learning rate = 0.001 

    with tf.name_scope('accuracy'): 
     correct = tf.equal(tf.argmax(prediction,1),tf.argmax(y,1)) 
     accuracy = tf.reduce_mean(tf.cast(correct,'float')) 
     tf.summary.scalar('accuracy',accuracy) 

    # cycles of feed forward and backprop 
    num_epochs = ne 


    with tf.Session() as sess: 
     saver = tf.train.Saver() 
     sess.run(tf.global_variables_initializer()) 
     merged_summary = tf.summary.merge_all() 

     for epoch in range(num_epochs): 
      epoch_loss = 0 
      for i in range(int(real_X_9.shape[0])/batch_size):#mnist.train.num_examples/batch_size)): # X.shape[0] 
       randidx = np.random.choice(real_X_9.shape[0], batch_size, replace=False) 
       epoch_x,epoch_y = real_X_9[randidx,:],real_y_9[randidx,:] #mnist.train.next_batch(batch_size) # X,y 
       j,c = sess.run([optimizer,cost],feed_dict={x:epoch_x,y:epoch_y,keep_prob:TRAIN_KEEP_PROB}) 
       if i == 0: 
        [ta] = sess.run([accuracy],feed_dict={x:epoch_x,y:epoch_y,keep_prob:TRAIN_KEEP_PROB}) 
        print 'Train Accuracy', ta 

       epoch_loss += c 
      print '\n','Epoch', epoch + 1, 'completed out of', num_epochs, '\nLoss:',epoch_loss 
     saver.save(sess, os.getcwd()+'/models/baseDNN') 
train(x) 

Voici mon code pour essayer de restaurer le modèle (qui est le code qui produit l'erreur):

import tensorflow as tf 
import os 

sess = tf.Session() 
saver = tf.train.import_meta_graph(os.getcwd()+'/models/baseDNN.meta') 
saver.restore(sess,tf.train.latest_checkpoint('./')) 

graph = tf.get_default_graph() 

x = graph.get_tensor_by_name('x_placeholder:0') 
y = graph.get_tensor_by_name('y_placeholder:0') 

op1 = graph.get_tensor_by_name('op1:0') 
op2 = graph.get_tensor_by_name('op2:0') 
op3 = graph.get_tensor_by_name('op3:0') 
op4 = graph.get_tensor_by_name('op4:0') 
op5 = graph.get_tensor_by_name('op5:0') 
op6 = graph.get_tensor_by_name('op6:0') 
op7 = graph.get_tensor_by_name('op7:0') 

Toute aide serait grandement apprécié. Merci.

Answer 1

Ne pensez pas que le fichier de point de contrôle correct est chargé. Vérifiez les fichiers enregistrés dans votre répertoire /models/baseDNN. Il devrait contenir model.ckpt et model.ckpt.meta (ou quelque chose de similaire). Pointez sur les fichiers appropriés:

saver = tf.train.import_meta_graph('/path/baseDNN.meta') 
saver.restore(sess,'/path/baseDNN')