Hello! I am trying to write a simple GRU decoder, such as depicted below <a ta

See my last answer at : <a class="issue-link js-issue-link" data-error-text="Failed to

I've just tried code from docs/readout.md <div cl

Error with initial_states while trying to write a GRU decoder about recurrentshop HOT 4 CLOSED

msurtsukov commented on August 22, 2024

Error with initial_states while trying to write a GRU decoder

from recurrentshop.

Comments (4)

farizrahman4u commented on August 22, 2024

See my last answer at : keras-team/keras#2968

By the way, thanks for this wonderful peace of software, which Keras lacks so much.

You are welcome!

from recurrentshop.

farizrahman4u commented on August 22, 2024

Also checkout docs/readout.md

from recurrentshop.

msurtsukov commented on August 22, 2024

I've just tried code from docs/readout.md

x = Input((10,))
h_tm1 = Input((10,))
c_tm1 = Input((10,))
readout_input = Input((10,))


# Here, I simply add half the readout to the input.. you can do whatever you want.

readout_half = Lambda(lambda x: 0.5 * x)(readout_input)
lstms_input = add([x, readout_half])

# Deep LSTM
depth = 3

cells = [LSTMCell(10) for _ in range(depth)]

lstms_output, h, c = lstms_input, h_tm1, c_tm1


for cell in cells:
    lstms_output, h, c = cell([lstms_output, h, c])

y = lstms_output

rnn = RecurrentModel(input=x, initial_states=[h_tm1, c_tm1], output=y, final_states=[h, c], readout_input=readout_input)

and ran into an error:

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-29-28879aa4ed9a> in <module>()
     23 y = lstms_output
     24 
---> 25 rnn = RecurrentModel(input=x, initial_states=[h_tm1, c_tm1], output=y, final_states=[h, c], readout_input=readout_input)

/Users/mikhail/anaconda/lib/python3.6/site-packages/recurrentshop-1.0.0-py3.6.egg/recurrentshop/engine.py in __init__(self, input, output, initial_states, final_states, readout_input, teacher_force, decode, output_length, return_states, **kwargs)
    281             raise Exception('Readout should be enabled for teacher forcing.')
    282         self.teacher_force = teacher_force
--> 283         self.model = Model(inputs, outputs)
    284         super(RecurrentModel, self).__init__(**kwargs)
    285         input_shape = list(K.int_shape(input))

/Users/mikhail/anaconda/lib/python3.6/site-packages/keras/legacy/interfaces.py in wrapper(*args, **kwargs)
     86                 warnings.warn('Update your `' + object_name +
     87                               '` call to the Keras 2 API: ' + signature, stacklevel=2)
---> 88             return func(*args, **kwargs)
     89         wrapper._legacy_support_signature = inspect.getargspec(func)
     90         return wrapper

/Users/mikhail/anaconda/lib/python3.6/site-packages/keras/engine/topology.py in __init__(self, inputs, outputs, name)
   1702                                 'The following previous layers '
   1703                                 'were accessed without issue: ' +
-> 1704                                 str(layers_with_complete_input))
   1705                     for x in node.output_tensors:
   1706                         computable_tensors.append(x)

RuntimeError: Graph disconnected: cannot obtain value for tensor Tensor("input_4:0", shape=(?, 10), dtype=float32) at layer "input_4". The following previous layers were accessed without issue: []

There are also some typos in docs: LSTM instead of LSTMCell and an issue with Lambda layer.

Also when using decode=True and setting output_length isn't it doing something like

cells = [LSTMCell(10) for _ in range(depth)]
lstms_output, h, c = lstms_input, h_tm1, c_tm1
for cell in cells:
    lstms_output, h, c = cell([lstms_output, h, c])

internally?

And could you, please, show how to run this rnn layer in the graph, I mean should it be done like this?

rnn(input, initial_states=[x], readout_input=readout)

Sorry for being annoying.

from recurrentshop.

farizrahman4u commented on August 22, 2024

and ran into an error:

Should be fixed now.

Also when using decode=True and setting output_length isn't it doing something like ... internally?

Sorry.. I don't get you.. can you elaborate??

And could you, please, show how to run this rnn layer in the graph

You use it like a regular Keras Recurrent layer.

y = rnn(x)

You can also provide initial states (default is zeros):

y = rnn(x, initial_state=[state1, state2..])

The initial readout (at t=0) will be zeros. You can set it using a tensor if you wish:

y = rnn(x, initial_state=[state1, state2..], initial_readout=some_tensor)

from recurrentshop.

Error with initial_states while trying to write a GRU decoder about recurrentshop HOT 4 CLOSED

Comments (4)

Related Issues (20)

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent