theano.gof.fg.MissingInputError: (......(keras_learning_phase) When i use the Dropout layer

[alyato]

hi guys. I meet the question when i load the predict function belongs to the graph

I know someone has issued this question But i still don't how to use K.learning_phase()

#2417

#2430

And if i remove the Dropout layer ,it work well.

thanks

[bottydim]

Hi @alyato, how did you fix your problem, I am running into the same issue after a GRU followed by dropout

[alyato]

when i call the model.predict ,it show me this error.

SO i load my model and recompile it. Then i call model.predict,it works

If you want to get the output after very layer and use the dropout layer,
you have to set the keras_learning_phase value 0/1.

2016-05-25 4:59 GMT-06:00 Botty Dimanov notifications@github.com:

Hi @alyato https://github.com/alyato, how did you fix your problem, I
am running into the same issue after a GRU followed by dropout

—
You are receiving this because you were mentioned.
Reply to this email directly or view it on GitHub
#2771 (comment)

[franciscovargas]

how do you recompile it ??? compile with the same arguments ? it does not sabotage the model being trained?

[alyato]

@franciscovargas first,you can save your model to yaml or json and save the weight to h5 format after calling the fit function.
then you need call the function model_from_json or model_from_yaml and load your weight when you want to call the predict function

[franciscovargas]

thanks so much !

On Tue, Jun 21, 2016 at 9:01 AM, alyato notifications@github.com wrote:

@franciscovargas https://github.com/franciscovargas first,you can save
your model to yaml or json and save the weight to h5 format after calling
the fit function.
then you need call the function model_from_json or model_from_yaml and
load your weight when you want to call the predict function

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#2771 (comment),
or mute the thread
https://github.com/notifications/unsubscribe/AFSJTEzTwK4qiUqwtIGBmtqfuMPgHrZzks5qN5p0gaJpZM4Ijs2q
.