axis error in recurrent.py

[WangYuxuan93]

Hi, I'm trying the parser training module with the basic python main.py train ParserNetwork command and only changed the directories in the default.cfg to my own. Then I got this error:

ValueError: Shape must be at least rank 3 but is rank 2 for 'ParserNetwork_1/Embeddings/form/Subtoken/Subtoken/RNN-0/RNN/concat' (op: 'ConcatV2') with input shapes: [100,400], [100,1200], [] and with computed input tensors: input[2] = <2>.

from /disk3/work/graph_based_parser/Parser-v3/parser/neural/recurrent.py", line 129, in LSTM

Where I found that the weights and gate_weights are loaded with shape of 2-dimension:

weights = tf.get_variable('Weights', shape=[input_size, recur_size])#, initializer=tf.orthogonal_initializer)
gate_weights = tf.get_variable('Gate_Weights', shape=[input_size, gate_size])#, initializer=tf.orthogonal_initializer)

While the concatenation are applied to dimension 3:

weights = tf.concat([weights, gate_weights], axis=2)

Thought this might be a mistake.

BTW, I didn't find the README in config directory mentioned in the README in root directory, did you forget to upload it?