ValueError: Graph disconnected: cannot obtain value for tensor Tensor

Question

I wrote a architecture similar to this code here without using sequential layer but it's returning ValueError,

# Input Memory Representation.
input_story = layers.Input(shape=(story_maxlen,), dtype='int32')
input_story_0 = layers.Embedding(input_dim=vocab_size, output_dim=64)(input_story)
input_story_1 = layers.Dropout(0.3)(input_story_0)

input_question = layers.Input(shape=(query_maxlen,), dtype='int32')
input_question_0 = layers.Embedding(input_dim=vocab_size, output_dim=64)(input_question)
input_question_1 = layers.Dropout(0.3)(input_question_0)

match = layers.dot([input_story_1, input_question_1], axes=(2, 2))
match = layers.Activation('softmax')(match)

# Output Memory Representation.
input_story_11 = layers.Input(shape=(story_maxlen,), dtype='int32')
input_story_12 = layers.Embedding(input_dim=vocab_size, output_dim=query_maxlen)(input_story_11)
input_story_13 = layers.Dropout(0.3)(input_story_12)

add = layers.add([match, input_story_13])
add = layers.Permute((2, 1))(add)

# Generating Final Predictions
x = layers.concatenate([add, input_question_1])
x = layers.LSTM(32)(x)
x = layers.Dropout(0.3)(x)
x = layers.Dense(vocab_size)(x)
x - layers.Activation('softmax')(x)

model = Model(inputs=[input_story, input_question], outputs=x)

This is the error I was getting

ValueError: Graph disconnected: cannot obtain value for tensor Tensor("input_143:0", shape=(None, 552), dtype=int32) at layer "input_143". The following previous layers were accessed without issue: ['input_142', 'input_141', 'embedding_141', 'embedding_140', 'dropout_152']

I've re-checked all the input layers, sizes and all seems to be fine but I can't figure out why I'm getting a value error. Can anyone help out?

Vivek Mehta · Accepted Answer · 2019-11-21 08:20:32Z

1

You have three Input layers in your code. You have added only two layers in your model input.

# Output Memory Representation.
input_story_11 = layers.Input(shape=(story_maxlen,), dtype='int32')

This layer is not added to model inputs. Hence it is giving graph disconnected error.

Add this layer to your inputs:

model = Model(inputs=[input_story, input_question, input_story_11], outputs=x)

answered Nov 21, 2019 at 8:20

Vivek Mehta

2,6542 gold badges20 silver badges32 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

user_12 Over a year ago

thank you but if you see the architecture from this post they have only given two layers in the model input. I was trying to replicate that. Can you check once.

Vivek Mehta Over a year ago

In this post I am seeing only two Input layers: input_sequence and question, while you have three.

user_12 Over a year ago

So am I doing something wrong here? I'm just trying to replicate that exact architecture but without using Sequential. Are there any mistakes that I need to correct to make both alike?

Vivek Mehta Over a year ago

@user_12 I didn't go through entire code, but you are passing third input input_story_13 to add layer, while in the blog it is different. You might want to start with that.

Collectives™ on Stack Overflow

ValueError: Graph disconnected: cannot obtain value for tensor Tensor

1 Answer 1

4 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

4 Comments

Your Answer

Sign up or log in

Post as a guest

Related