1
回答
ValueError: Attempt to have a second RNNCell use the weights of a variable scope that already has weights:
科大讯飞通用文字识别100000次/天免费使用。立即申请   

@黄文坚 你好,想跟你请教个问题:

我跑了下了您书上 关于ptb的lstm程序,出现了 这样的错误,好像是说weight被多次赋值了,我寻思了许久,始终找不到错误。

ValueError: Attempt to have a second RNNCell use the weights of a variable scope that already has weights: 'Model/RNN/multi_rnn_cell/cell_0/basic_lstm_cell'; and the cell was not constructed as BasicLSTMCell(..., reuse=True).  To share the weights of an RNNCell, simply reuse it in your second calculation, or create a new one with the argument reuse=True.
 

我的代码:

import  time
import tensorflow as tf
import numpy as np
import reader

class PTBInput(object):

    def __init__(self,config,data,name = None):
        self.batch_size = batch_size = config.batch_size
        self.num_steps = num_steps = config.num_steps
        self.epoch_size = ((len(data) // batch_size) -1) // num_steps
        self.input_data,self.targets = reader.ptb_producer(data,batch_size,num_steps,name = name)


class PTBModel(object):

    def __init__(self,is_training,config,input_):
        self._input = input_

        batch_size = input_.batch_size
        num_steps = input_.num_steps
        size = config.hidden_size
        vocab_size = config.vocab_size

        def lstm_cell():
            return tf.contrib.rnn.BasicLSTMCell(
                size,forget_bias = 0.0,state_is_tuple = True)
        attn_cell = lstm_cell
        if is_training and config.keep_prob < 1:
            def attn_cell():
                return tf.contrib.rnn.DropoutWrapper(
                    lstm_cell(),output_keep_prob=config.keep_prob)
        cell = tf.contrib.rnn.MultiRNNCell(
            [attn_cell() for _ in range(config.num_layers)],
            state_is_tuple = True)

        self._initial_state = cell.zero_state(batch_size,tf.float32)

        with tf.device("/cpu:0"):
            embedding = tf.get_variable(
                "embedding",[vocab_size,size],dtype = tf.float32 )
            inputs = tf.nn.embedding_lookup(embedding ,input_.input_data)

        if is_training and config.keep_prob < 1:
            inputs = tf.nn.dropout(inputs,config.keep_prob)

        outputs = []
        state = self._initial_state

        with tf.variable_scope("RNN"):
            for time_step in range(num_steps):
                if time_step > 0: tf.get_variable_scope().reuse_variables()
                (cell_output,state) = cell(inputs[:, time_step,:],state)
                outputs.append(cell_output)

        output = tf.reshape(tf.concat(outputs,1),[-1,size])
        softmax_w = tf.get_variable(
            "softmax_w",[size,vocab_size],dtype=tf.float32)
        softmax_b = tf.get_variable("softmax_b",[vocab_size],dtype=tf.float32)
        logits = tf.matmul(output,softmax_w) + softmax_b
        loss = tf.contrib.legacy_seq2seq.sequence_loss_by_example(
            [logits],
            [tf.reshape(input_.targets,[-1])],
            [tf.ones([batch_size * num_steps],dtype=tf.float32)]
        )
        self._cost = cost = tf.reduce_sum(loss) / batch_size
        self._final_state = state

        if not is_training:
            return

        self._lr = tf.Variable(0.0,trainable=False)
        tvars = tf.trainable_variables()
        grads,_ = tf.clip_by_global_norm(tf.gradients(cost,tvars),
                                         config.max_grad_norm)
        optimizer = tf.train.GradientDescentOptimizer(self._lr)
        self._train_op = optimizer.apply_gradients(zip(grads,tvars),
        global_step=tf.contrib.framework.get_or_create_global_step()
                            )

        self._new_lr = tf.placeholder(
            tf.float32,shape= [],name="new_learning_rate"  )
        self._lr_update = tf.assign(self._lr,self._new_lr)

    def assign_lr(self,session,lr_value):
        session.run(self._lr_update,feed_dict = {self._new_lr:lr_value})

    @property
    def input(self):
        return self._input

    @property
    def initial_state(self):
        return self._initial_state

    @property
    def cost(self):
        return self._cost

    @property
    def final_state(self):
        return self.final_state

    @property
    def lr(self):
        return self._lr

    @property
    def train_op(self):
        return self._train_op

class SmallConfig(object):
    init_scal = 0.1
    learning_rate = 1.0
    max_grad_norm = 5
    num_layers = 2
    num_steps =20
    hidden_size = 200
    max_epoch = 4
    max_max_epoch = 13
    keep_prob = 1.0
    lr_decay = 0.5
    batch_size = 20
    vocab_size = 10000

class MediumConfig(object):
    init_scale = 0.05
    learning_rate = 1.0
    max_grad_norm = 5
    num_layers = 2
    num_steps = 35
    hidden_size = 650
    max_epoch = 6
    max_max_epoch = 39
    keep_prob = 0.5
    lr_decay = 0.8
    batch_size = 20
    vocab_size = 10000

class LargeConfig(object):
    init_scal = 0.04
    learing_rate = 1.0
    max_grad_norm = 10
    num_layers =2
    num_steps = 35
    hidden_size = 1500
    max_epoch =14
    max_max_epoch = 55
    keep_prob = 0.35
    lr_decay = 1/1.55
    batch_size = 20
    vocab_size = 10000

class TestConfig(object):
    init_scale =0.1
    learning_rate = 1.0
    max_grad_norm =1
    num_layers =1
    num_step2 =2
    hidden_size = 2
    max_epoch = 1
    max_max_epoch =1
    keep_prob = 1.0
    lr_decay = 0.5
    batch_size = 20
    vocab_size  = 10000

def run_epoch(session,model,eval_op = None,verbose=False):
    start_time = time.time()
    costs = 0.0
    iters = 0
    state = session.run(model.initial_state)

    fetches = {
        "cost":model.cost,
        "final_state":model.final_state,
               }
    if eval_op is not None:
        fetches["eval_op"] = eval_op

    for step in range(model.input.epoch_size):
        feed_dict = {}
        for i,(c,h) in enumerate(model.initial_state):
            feed_dict[c] = state[i].c
            feed_dict[h] = state[i].h

        vals = session.run(fetches,feed_dict)
        cost = vals["cost"]
        state = vals["final_state"]

        cost+=cost
        iters += model.input.num_steps
        if verbose and step % (model.input.epoch_size // 10) == 10:
            print ("%.3f perplexity:%.3f speed: %.0f wps"
                   % (step * 1.0 / model.input.epoch_size,np.exp(costs / iters),
                      iters * model.input.batch_size / (time.time() - start_time)))
    return np.exp(costs / iters)

raw_data = reader.ptb_raw_data('simple-examples/data/')
train_data,valid_data,test_dat,_ = raw_data

config = SmallConfig()
eval_config = SmallConfig()
eval_config.batch_size = 1
eval_config.num_steps = 1

with tf.Graph().as_default():
    initializer = tf.random_uniform_initializer(-config.init_scal,
                                                config.init_scal)

    with tf.name_scope("Train"):
        train_input = PTBInput(config = config ,data =train_data,name = "TrainInput")
        with tf.variable_scope("Model",reuse = None,initializer = initializer):
            m = PTBModel(is_training = True,config = config,input_ = train_input)

    with tf.name_scope("Valid"):
        valid_input = PTBInput(config = config,data = valid_data,name = "ValidInput")
        with tf.variable_scope("Model",reuse = True, initializer = initializer):
            mvalid = PTBModel(is_training = False,config = config,input_ = valid_input)

    with tf.name_scope("Test"):
        test_input = PTBInput(config = eval_config,data = test_dat,name = "TestInput")
        with tf.variable_scope("Model",reuse = True,initializer = initializer):
            mtest = PTBModel(is_training=False,config = eval_config,input_ = test_input)

    sv = tf.train.Supervisor()
    with sv.managed_session() as session:
        for i in range(config.max_max_epoch):
            lr_decay = config.lr_decay ** max(i + 1 - config.max_epoch , 0.0)
            m.assign_lr(session,config.learning_rate * lr_decay)

            print("Epoch: %d Learning rate: %.3f" % (i+1,session.run(m.lr)))
            train_perplexity = run_epoch(session,m,eval_op = m.tain_op,verbose=True)
            print("Epoch: %d Train Perplexity: %.3f" % (i + 1, session.run(train_perplexity)))
            valid_perplexity = run_epoch(session, mvalid)
            print("Epoch: %d Valid Perplexity: %.3f" % (i+1, valid_perplexity))

        test_perplexity = run_epoch(session,mtest)
        print("Test Perplexity: %.3f" % test_perplexity)
<无标签>
举报
xiaowei201314
发帖于1年前 1回/1K+阅
顶部