Theano教程：UnusedInputError：theano.function

Question

Attempting to teach myself neural nets, I have begun working my way through the Theano tutorials over at deeplearning.net. I just encountered an error that I was not expecting as I have literally copied and pasted every line of code from the tutorial. I'm sure whatever is wrong is something small and I'm just looking over it but any help would be much appreciated. Thanks

http://deeplearning.net/software/theano/tutorial/examples.html#copying-functions

import theano
import theano.tensor as T
state = theano.shared(0)
inc = T.iscalar('inc')
accumulator = theano.function([inc], state, updates=[(state, state+inc)])
accumulator(10)
print(state.get_value())

new_state = theano.shared(0)
new_accumulator = accumulator.copy(swap={state:new_state})
new_accumulator(100)

print(state.get_value())
print(new_state.get_value())

null_accumulator = accumulator.copy(delete_updates=True)


---------------------------------------------------------------------------
UnusedInputError                          Traceback (most recent call last)
<ipython-input-20-5d1acb597345> in <module>()
----> 1 null_accumulator = accumulator.copy(delete_updates=True)

/home/mcamp/anaconda3/lib/python3.5/site-packages/theano/compile/function_module.py in copy(self, share_memory, swap, delete_updates, name, profile)
    719                                 # can contain inplace. DebugMode check
    720                                 # that.
--> 721                                 accept_inplace=True,
    722                                 ).create(input_storage,
    723                                          storage_map=new_storage_map)

/home/mcamp/anaconda3/lib/python3.5/site-packages/theano/compile/function_module.py in __init__(self, inputs, outputs, mode, accept_inplace, function_builder, profile, on_unused_input, fgraph, output_keys)
   1413 
   1414         # Check if some input variables are unused
-> 1415         self._check_unused_inputs(inputs, outputs, on_unused_input)
   1416 
   1417         # Make a list of (SymbolicInput|SymblicInputKits, indices,

/home/mcamp/anaconda3/lib/python3.5/site-packages/theano/compile/function_module.py in _check_unused_inputs(self, inputs, outputs, on_unused_input)
   1551                 elif on_unused_input == 'raise':
   1552                     raise UnusedInputError(msg % (inputs.index(i),
-> 1553                                                   i.variable, err_msg))
   1554                 else:
   1555                     raise ValueError("Invalid value for keyword "

UnusedInputError: theano.function was asked to create a function computing outputs given certain inputs, but the provided input variable at index 0 is not part of the computational graph needed to compute the outputs: inc.
To make this error into a warning, you can pass the parameter on_unused_input='warn' to theano.function. To disable it completely, use on_unused_input='ignore'.

Answer 1

I'm also new at theano, and I just ran into this. It seems to be a bug in the tutorial. The error message suggests using the parameter on_unused_input to suppress the error, but this isn't a keyword argument for copy . Apparently it has to be passed to function . So if accumulator was created like this:

accumulator = theano.function([inc], state,
    updates=[(state, state+inc)], on_unused_input='ignore')

then the error goes away. (This seems to me like a non-ideal solution, having to change a property of the original function.)