I am working on a text generalization network and must implement an encoder for use with tf.nn.seq2seq.embedding_attention_decoder . As part of this, I need to encode various batches of sequences into representing vectors, but the internal encoding fails.
Here's a simplified snippet that gives the same error:
import tensorflow as tf single_cell = tf.nn.rnn_cell.GRUCell(1024) sentence_cell = tf.nn.rnn_cell.EmbeddingWrapper(single_cell, embedding_classes = 40000) batch = [tf.placeholder(tf.int32, [1,1]) for _ in range(250)] (_ , state) = tf.nn.rnn(sentence_cell, batch, dtype= tf.int32)
This fails with the following stack trace:
Traceback (most recent call last): File "/home/ubuntu/workspace/example.py", line 6, in <module> (_ , state) = tf.nn.rnn(sentence_cell, batch, dtype= tf.int32) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/rnn.py", line 126, in rnn (output, state) = call_cell() File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/rnn.py", line 119, in <lambda> call_cell = lambda: cell(input_, state) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/rnn_cell.py", line 616, in __call__ return self._cell(embedded, state) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/rnn_cell.py", line 150, in __call__ 2 * self._num_units, True, 1.0)) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/rnn_cell.py", line 706, in linear res = math_ops.matmul(array_ops.concat(1, args), matrix) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/array_ops.py", line 314, in concat name=name) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/gen_array_ops.py", line 70, in _concat name=name) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/op_def_library.py", line 396, in apply_op raise TypeError("%s that don't all match." % prefix) TypeError: Tensors in list passed to 'values' of 'Concat' Op have types [float32, int32] that don't all match.
When debugging the input size of sentence_cell is 1 and the elements in batch are of size [1,1] , which is infact [batch_size, sentence_cell.input_size] .
Switching to dtype = tf.float32 in the call to tf.nn.rnn() does the job snippet, but gives me the following stack trace in my code:
[nltk_data] Downloading package punkt to /home/alex/nltk_data... [nltk_data] Package punkt is already up-to-date! Preparing news data in . Creating 3 layers of 1024 units. > /home/alex/Programmering/kandidatarbete/arbete/code/seq3seq/seq3seq_model.py(84)encode_sentence() -> (_ ,state) = tf.nn.rnn(sentence_cell, sent, sequence_length = length, dtype= tf.float32) (Pdb) c Traceback (most recent call last): File "translate.py", line 268, in <module> tf.app.run() File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/default/_app.py", line 30, in run sys.exit(main(sys.argv)) File "translate.py", line 265, in main train() File "translate.py", line 161, in train model = create_model(sess, False) File "translate.py", line 136, in create_model forward_only=forward_only) File "/home/alex/Programmering/kandidatarbete/arbete/code/seq3seq/seq3seq_model.py", line 141, in __init__ softmax_loss_function=None) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/seq2seq.py", line 926, in model_with_buckets decoder_inputs[:bucket[1]]) File "/home/alex/Programmering/kandidatarbete/arbete/code/seq3seq/seq3seq_model.py", line 140, in <lambda> lambda x, y: seq3seq_f(x, y, False), File "/home/alex/Programmering/kandidatarbete/arbete/code/seq3seq/seq3seq_model.py", line 98, in seq3seq_f art_vecs = tfmap(encode_article, tf.pack(encoder_inputs)) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 1900, in map _, r_a = While(lambda i, a: math_ops.less(i, n), compute, [i, acc_ta]) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 1557, in While result = context.BuildLoop(cond, body, loop_vars) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 1474, in BuildLoop body_result = body(*vars_for_body_with_tensor_arrays) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 1897, in compute a = a.write(i, fn(elems_ta.read(i))) File "/home/alex/Programmering/kandidatarbete/arbete/code/seq3seq/seq3seq_model.py", line 92, in encode_article return tfmap(encode_sentence, article) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 1900, in map _, r_a = While(lambda i, a: math_ops.less(i, n), compute, [i, acc_ta]) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 1557, in While result = context.BuildLoop(cond, body, loop_vars) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 1474, in BuildLoop body_result = body(*vars_for_body_with_tensor_arrays) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 1897, in compute a = a.write(i, fn(elems_ta.read(i))) File "/home/alex/Programmering/kandidatarbete/arbete/code/seq3seq/seq3seq_model.py", line 84, in encode_sentence (_ ,state) = tf.nn.rnn(sentence_cell, sent, sequence_length = length, dtype= tf.float32) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/rnn.py", line 124, in rnn zero_output, state, call_cell) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/rnn.py", line 212, in _rnn_step time < max_sequence_length, call_cell, empty_update) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 1183, in cond res_t = context_t.BuildCondBranch(fn1) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 1106, in BuildCondBranch r = fn() File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/rnn.py", line 119, in <lambda> call_cell = lambda: cell(input_, state) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/rnn_cell.py", line 615, in __call__ embedding, array_ops.reshape(inputs, [-1])) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/embedding_ops.py", line 86, in embedding_lookup validate_indices=validate_indices) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/gen_array_ops.py", line 423, in gather validate_indices=validate_indices, name=name) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/op_def_library.py", line 486, in apply_op _Attr(op_def, input_arg.type_attr)) File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/op_def_library.py", line 59, in _SatisfiesTypeConstraint ", ".join(dtypes.as_dtype(x).name for x in allowed_list))) TypeError: DataType float32 for attr 'Tindices' not in list of allowed values: int32, int64
What am I missing?