Error : An input of the graph, used to compute Elemwise{Cast{float64}}(mask_1), was not provide

11 views
Skip to first unread message

Satyan Sharma

unread,
Nov 5, 2019, 7:36:07 AM11/5/19
to theano-users
Hi Theano users,

I am noew to Theano and been trying to work on a code from git hub repo  https://github.com/CancerDataScience/SurvivalNet.

I have been trying to reproduce the work, but not sure where I am missing. I can get the optimized model but I get stuck at the model ananlysis part.


partial_derivative = th.function(on_unused_input='ignore',
inputs=[X, AtRisk, Observed, Is_train, masks],
outputs=T.grad(Model.risk_layer.output[0],
Model.x),
givens={Model.x: X, Model.o: AtRisk,
Model.at_risk: Observed,
Model.is_train: Is_train,
Model.masks[0]: masks},
name='partial_derivative')

Runing the

Satyan Sharma

unread,
Nov 5, 2019, 7:39:58 AM11/5/19
to theano-users
Hello again,

Sorry I accidentally got it posted, while I was still continuing to write.
Runing the partial derivative block seems to raise the error 

theano.gof.fg.MissingInputError: ("An input of the graph, used to compute Elemwise{Cast{float64}}(mask_1), was not provided and not given a value.Use the Theano flag exception_verbosity='high',for more information on this error.", mask_1)

Starngely, It works when the network a single layer. Upon increasing the number of layers, the error shows up.
I am using Theano (0.8.2).

Any help would be greatly appreciated. 
best regards,
Satyan
Reply all
Reply to author
Forward
0 new messages