我构建了一个LSTM网络，我的输入维度是100*100*83 (batch_size=100, steps = 100, char_vector = 83)。我构建了两个具有512个隐藏单元的LSTM层。
# coding: utf-8from __future__ import print_functionimport tensorflow as tfimport numpy as npimport timeclass CharRNN: def __init__(self, num_classes, batch_size=64, num_steps=50, lstm_size=128, num_layers =2,\ learning_rate = 0.001, grad_clip=5, keep_prob=0.001,sampling= False): # True for SGD if sampling == True: self.batch_size, self.num_steps = 1,1 else: self.batch_size, self.num_steps = batch_size, num_steps tf.reset_default_graph() self.inputs, self.targets, self.keep_prob = self.build_inputs(self.batch_size,self.num_steps) self.keep_prob = keep_prob self.cell, self.initial_state = self.build_lstm(lstm_size,num_layers,self.batch_size,self.keep_prob) # print(self.cell.state_size) x_one_hot = tf.one_hot(self.inputs, num_classes) print("cell state size: ",self.cell.state_size) print("cell initial state: ",self.initial_state) print("this is inputs", self.inputs) print("x_one_hot: ",x_one_hot) outputs, state = tf.nn.dynamic_rnn(self.cell, x_one_hot, initial_state= self.initial_state) def build_inputs(self, num_seqs, num_steps): inputs = tf.placeholder(tf.int32, shape=(num_seqs, num_steps), name = "inputs") targets = tf.placeholder(tf.int32, shape= (num_seqs, num_steps), name="targets") print('inputs shape: ',inputs.shape) keep_prob = tf.placeholder(tf.float32, name="keep_prob") return inputs, targets, keep_prob def build_lstm(self, lstm_size, num_layers, batch_size, keep_prob): # construct lstm cell lstm = tf.nn.rnn_cell.BasicLSTMCell(lstm_size) # add dropout drop = tf.nn.rnn_cell.DropoutWrapper(lstm, output_keep_prob= keep_prob) # stack multiple rnn cells cell = tf.nn.rnn_cell.MultiRNNCell([drop for _ in range(num_layers)]) initial_state = cell.zero_state(batch_size, tf.float32) return cell, initial_stateif __name__ == '__main__': len_vocab = 83 batch_size = 100 num_steps = 100 lstm_size = 512 num_layers = 2 learning_rate = 0.001 keep_prob = 0.5 epochs = 20 save_every_n = 200 print("h1") model = CharRNN(len_vocab, batch_size = batch_size, num_steps=num_steps, lstm_size = lstm_size,num_layers=num_layers\ ,learning_rate=learning_rate,sampling= False,keep_prob = keep_prob我在tf.nn.dynamic_rnn处遇到了维度不匹配错误。错误信息如下：
inputs shape: (100, 100)cell state size: (LSTMStateTuple(c=512, h=512), LSTMStateTuple(c=512, h=512))cell initial state: (LSTMStateTuple(c=<tf.Tensor 'MultiRNNCellZeroState/DropoutWrapperZeroState/BasicLSTMCellZeroState/zeros:0' shape=(100, 512) dtype=float32>, h=<tf.Tensor 'MultiRNNCellZeroState/DropoutWrapperZeroState/BasicLSTMCellZeroState/zeros_1:0' shape=(100, 512) dtype=float32>), LSTMStateTuple(c=<tf.Tensor 'MultiRNNCellZeroState/DropoutWrapperZeroState_1/BasicLSTMCellZeroState/zeros:0' shape=(100, 512) dtype=float32>, h=<tf.Tensor 'MultiRNNCellZeroState/DropoutWrapperZeroState_1/BasicLSTMCellZeroState/zeros_1:0' shape=(100, 512) dtype=float32>))this is inputs Tensor("inputs:0", shape=(100, 100), dtype=int32)x_one_hot: Tensor("one_hot:0", shape=(100, 100, 83), dtype=float32)Traceback (most recent call last): File "./seq2_minimal.py", line 70, in <module> ,learning_rate=learning_rate,sampling= False,keep_prob = keep_prob) File "./seq2_minimal.py", line 32, in __init__ outputs, state = tf.nn.dynamic_rnn(self.cell, x_one_hot, initial_state= self.initial_state) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/rnn.py", line 614, in dynamic_rnn dtype=dtype) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/rnn.py", line 777, in _dynamic_rnn_loop swap_memory=swap_memory) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/control_flow_ops.py", line 2816, in while_loop result = loop_context.BuildLoop(cond, body, loop_vars, shape_invariants) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/control_flow_ops.py", line 2640, in BuildLoop pred, body, original_loop_vars, loop_vars, shape_invariants) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/control_flow_ops.py", line 2590, in _BuildLoop body_result = body(*packed_vars_for_body) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/rnn.py", line 762, in _time_step (output, new_state) = call_cell() File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/rnn.py", line 748, in <lambda> call_cell = lambda: cell(input_t, state) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/rnn_cell_impl.py", line 183, in __call__ return super(RNNCell, self).__call__(inputs, state) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/layers/base.py", line 575, in __call__ outputs = self.call(inputs, *args, **kwargs) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/rnn_cell_impl.py", line 1066, in call cur_inp, new_state = cell(cur_inp, cur_state) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/rnn_cell_impl.py", line 891, in __call__ output, new_state = self._cell(inputs, state, scope) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/rnn_cell_impl.py", line 183, in __call__ return super(RNNCell, self).__call__(inputs, state) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/layers/base.py", line 575, in __call__ outputs = self.call(inputs, *args, **kwargs) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/rnn_cell_impl.py", line 441, in call value=self._linear([inputs, h]), num_or_size_splits=4, axis=1) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/rnn_cell_impl.py", line 1189, in __call__ res = math_ops.matmul(array_ops.concat(args, 1), self._weights) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/math_ops.py", line 1891, in matmul a, b, transpose_a=transpose_a, transpose_b=transpose_b, name=name) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/ops/gen_math_ops.py", line 2437, in _mat_mul name=name) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 787, in _apply_op_helper op_def=op_def) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2958, in create_op set_shapes_for_outputs(ret) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2209, in set_shapes_for_outputs shapes = shape_func(op) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2159, in call_with_requiring return call_cpp_shape_fn(op, require_shape_fn=True) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/framework/common_shapes.py", line 627, in call_cpp_shape_fn require_shape_fn) File "/usr/local/lib/python2.7/site-packages/tensorflow/python/framework/common_shapes.py", line 691, in _call_cpp_shape_fn_impl raise ValueError(err.message)ValueError: Dimensions must be equal, but are 1024 and 595 for 'rnn/while/rnn/multi_rnn_cell/cell_0/cell_0/basic_lstm_cell/MatMul_1' (op: 'MatMul') with input shapes: [100,1024], [595,2048].我搜索后发现tensorflow的lstm单元应该能自动调整其输入大小。但错误信息显示了相反的情况。
它显示
输入大小为[100, 1024]，而lstm为[595, 2048]。
首先感谢您。
回答：
cell = tf.nn.rnn_cell.MultiRNNCell([drop for _ in range(num_layers)])
改为
cell = tf.nn.rnn_cell.MultiRNNCell([drop])
因为您给定的输入张量和生成的张量不一致。
