Config parameter settings in 02_02Session

tensorflow TF AS Import 
Import numpy NP AS 

# TODO learning Session Config = parameters used tf.ConfigProto () a. GPU focus parameter related 


DEF config_params (): 
    # TODO second way of performing the session 
    with tf.Graph () as_default ():. 
        # A, building the model in FIG 
        print ( 'default diagram of the current model is: {}' .format (tf.get_default_graph ())) 
        #. 1, the original definition of two input objects tensor 
        a tf.constant = ( 
            value = [1,2,3,4,5,6,3,4,3,45 ,. 5], DTYPE = tf.float32, Shape = [. 3,. 5], name = 'A' 
        ) 
        B tf.constant = ( 
            value = [3,3,3,3,3,3234,56,324,3,5 ], DTYPE = tf.float32, Shape = [. 5,. 3] 
        ) 

        # 2, using the above-described two op add constants were added a random number 
        v1 = a + np.random.random_sample ()
        tf.add = V2 (B, tf.random_normal (Shape = [], DTYPE = tf.float32, 43 is SEED =)) 

        #. 3, using two tensor matrix multiplication. 
        = tf.matmul rezult (V1, V2) 
        Print (A, B, V1, V2, rezult) 

        # two, building session 
        "" " 
        tf.Session ( 
            target = '', a url given connection, only the operation of the distributed the time required for a given 
            graph = None, Photo call which, if not given, it calls the default map 
            config configuration protocol = None) session. ""
        " 
        Optimizer = tf.OptimizerOptions ( 
            do_common_subexpression_elimination = True, # represents the open public execution sub optimization sentence 
            do_constant_folding = True, # set to True open showing constant folding optimization. 
            opt_level = 0 # 0 is set to indicate on the above two optimization, the default is 0 
        )
        = tf.GraphOptions graph_options (optimizer_options = Optimizer) 
        with tf.Session (config = tf.ConfigProto (graph_options = graph_options)) AS Sess: 
            Print (sess.run ([rezult, V2])) 


DEF config_params1 (): 
    # Here TODO increased equipment tf.device specified operation or create a variable. 
    tf.Graph with () as_default ():. 
        # a, building the model in FIG 
        print (. 'FIG current default model is: {}' the format (tf.get_default_graph ())) 
        #. 1, define two of the original input Object tensor 
        with tf.device ( '/ the GPU: 0'): 
            # A and b two constants will gpu: 0 created. 
            tf.constant = A ( 
                value = [1,2,3,4,5,6,3,4,3,45,5], DTYPE = tf.float32, Shape = [. 3,. 5], name = 'A ' 
            ) 
            B = tf.constant (
                = value [3,3,3,3,3,3234,56,324,3,5], DTYPE = tf.float32, Shape = [. 5,. 3] 
            ) 

        # 2, using the above-described two op add a constant by adding random number 
        V1 = A + np.random.random_sample () 
        V2 = tf.add (B, tf.random_normal (Shape = [], DTYPE = tf.float32, 43 is SEED =)) 

        #. 3, the use of two-phase matrix tensor multiply. 
        tf.device with ( '/ the GPU:. 1'): 
            rezult tf.matmul = (V1, V2) 
        Print (A, B, V1, V2, rezult) 

        # two, building session 
        "" " 
        tf.Session ( 
            target = ' ', to the url given connection, only time distributed operating needs of a given 
            graph = None, Photo call which, if not given, it calls the default map 
            config configuration protocol = None) session. 
        "" "
        tf.Session with (config = tf.ConfigProto (log_device_placement = True, 
                                              allow_soft_placement = True)) AS sess: 
            Print (sess.run ([rezult, v2])) 


DEF config_params2 (): 
    # TODO parameters related to learning gpu. 
    tf.Graph with () as_default ():. 
        # a, building the model in FIG 
        print (. 'FIG current default model is: {}' the format (tf.get_default_graph ())) 
        #. 1, define two of the original input Object tensor 
        with tf.device ( '/ the GPU: 0'): 
            # A and b two constants will gpu: 0 created. 
            tf.constant = A ( 
                value = [1,2,3,4,5,6,3,4,3,45,5], DTYPE = tf.float32, Shape = [. 3,. 5], name = 'A ' 
            ) 
            B = tf.constant (
                = value [3,3,3,3,3,3234,56,324,3,5], DTYPE = tf.float32, Shape = [. 5,. 3] 
            ) 

        # 2, using the above-described two op add a constant by adding random number 
        V1 = A + np.random.random_sample () 
        V2 = tf.add (B, tf.random_normal (Shape = [], DTYPE = tf.float32, 43 is SEED =)) 

        #. 3, the use of two-phase matrix tensor multiply. 
        tf.device with ( '/ the GPU:. 1'): 
            rezult tf.matmul = (V1, V2) 
        Print (A, B, V1, V2, rezult) 

        # two, building session 
        "" " 
        gpu_options parameters introduced 
           log_device_placement bool value the printing apparatus whether the log file at 
           allow_soft_placement bool value is used to allow dynamic tf cpu and gpu default False 
        "" " 
        gpu_options tf.GraphOptions = (
            allow_growth = True, # does not use the entire pre-allocated memory gpu calculation, but from small to large demand increase 
            per_process_gpu_memory_fraction = 0.8 # value range (0, 1), limit the use of the percentage of the memory device gpu. 
        ) 
        With tf.Session (config = tf.ConfigProto (log_device_placement = True, 
                                              allow_soft_placement = True, 
                                              gpu_options = gpu_options)) AS Sess: 
            Print (sess.run ([rezult, V2])) 


IF the __name__ == '__main__': 
    config_params1 ()