[theano-users] need to know what time_step i'm at in step, during scan?

Stephanie Peterson

2017-09-12 20:35:39 UTC

Quick follow up: I've included the two step functions above because
conditionals determine if step or step_masked is passed as the fn argument
to theano.scan, i.e. I need to add the time_step argument to both of them.
Also, I looked at stackoverflow's
theano-pass-iterating-index-of-scan-to-the-called-function
<https://stackoverflow.com/questions/41209427/theano-pass-iterating-index-of-scan-to-the-called-function>
in addition to the theano-users forum post I linked above.

Post by Stephanie Peterson
Hi all,
I need to know the current time step/iteration number within the step
function during scan for a custom GRU model. I have based my code to do
this on this theano-users thread
<https://groups.google.com/forum/#!topic/theano-users/Dd7okQ_U8Qg>,
however I am new to theano and am clearly missing something.
The relevant code snippets are below, along with the current error I am
receiving when I try to train my model (I've bolded what I have added to
the original model). Please let me know if I need to provide any other
context! Thanks in advance!
...
...
# some conditionals involving idx
...
hid = step(*idx*, input_n, hid_previous, *args)
...
# scan op iterated over first dimension of input and repeatedly
applies the step function
* idx = T.scalar('idx', dtype='int64')*
* sequences.insert(0, T.arange(idx.shape[0])) **# how do I
properly pass theano.tensor.arange as one of the sequences to the scan?*
* non_seqs.insert(0, idx)*
hid_out = theano.scan(
fn=step,
sequences=sequences,
go_backwards=self.backwards,
non_sequences = non_seqs,
truncate_gradient=self.gradient_steps,
strict=True,)[0]
...
...
* idx = T.scalar('idx', dtype='int64') **# should I create a shared
variable here instead?*
train_fn = theano.function([*idx*, in_x1, in_mask1, in_x2, in_mask2,
in_x3, in_mask3, in_y],
loss, updates=updates,
on_unused_input='warn')
File "main.py", line 362, in <module>
main(args)
File "main.py", line 220, in main
train_fn, test_fn, params, all_params = build_fn(args, embeddings)
File "main.py", line 154, in build_fn
test_fn = theano.function([idx, in_x1, in_mask1, in_x2, in_mask2,
in_x3, in_mask3, in_y], [acc, test_prediction], on_unused_input='warn')
File "/usr/local/lib/python2.7/dist-packages/theano/compile/function.py",
line 326, in function
output_keys=output_keys)
File "/usr/local/lib/python2.7/dist-packages/theano/compile/pfunc.py",
line 486, in pfunc
output_keys=output_keys)
File "/usr/local/lib/python2.7/dist-packages/theano/compile/function_module.py",
line 1794, in orig_function
output_keys=output_keys).create(
File "/usr/local/lib/python2.7/dist-packages/theano/compile/function_module.py",
line 1446, in __init__
accept_inplace)
File "/usr/local/lib/python2.7/dist-packages/theano/compile/function_module.py",
line 177, in std_fgraph
update_mapping=update_mapping)
File "/usr/local/lib/python2.7/dist-packages/theano/gof/fg.py", line
180, in __init__
self.__import_r__(output, reason="init")
File "/usr/local/lib/python2.7/dist-packages/theano/gof/fg.py", line
351, in __import_r__
self.__import__(variable.owner, reason=reason)
File "/usr/local/lib/python2.7/dist-packages/theano/gof/fg.py", line
397, in __import__
raise MissingInputError(error_msg, variable=r)
theano.gof.fg.MissingInputError: Input 0 of the graph (indices start from
0), used to compute Shape(idx), was not provided and not given a value. Use
the Theano flag exception_verbosity='high', for more information on this
error.
--
---
You received this message because you are subscribed to a topic in the
Google Groups "theano-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/
topic/theano-users/bP3ePKLSqd8/unsubscribe.
To unsubscribe from this group and all its topics, send an email to
For more options, visit https://groups.google.com/d/optout.

--
---
You received this message because you are subscribed to the Google Groups "theano-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to theano-users+***@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.