tensorflow tf.cond does not execute true_fn or false_fn for tf.reduce_mean - tensorflow

Question

I am trying to condition the output of the loss function tf.reduce_mean so as to avoid NaN errors. My code is:
limit=[]
for i in xrange(12):
limit.append(10000.0)
limit = tf.constant(limit)
predictions["loss"] =tf.cond(tf.reduce_mean(
(prediction - transformed_values) ** 2, axis=-1) < limit,
lambda:tf.reduce_mean(
(prediction - transformed_values) ** 2, axis=-1),
lambda:tf.reduce_mean(
(prediction - transformed_values), axis=-1)).
However, I get the error
INFO:tensorflow:Using default config.
WARNING:tensorflow:Using temporary folder as model directory: /tmp/tmpfnvr6j
INFO:tensorflow:Using config: {'_save_checkpoints_secs': 600, '_session_config': None, '_keep_checkpoint_max': 5, '_task_type': 'worker', '_is_chief': True, '_cluster_spec': <tensorflow.python.training.server_lib.ClusterSpec object at 0x7f7eaa5bd750>, '_save_checkpoints_steps': None, '_keep_checkpoint_every_n_hours': 10000, '_service': None, '_num_ps_replicas': 0, '_tf_random_seed': None, '_master': '', '_num_worker_replicas': 1, '_task_id': 0, '_log_step_count_steps': 100, '_model_dir': '/tmp/tmpfnvr6j', '_save_summary_steps': 100}
shape: pred (12,) true_t (12,) false_t (12,)
Traceback (most recent call last):
File "/home/paul/workspace/workspace/Master/Elec_Price_Prediction/Time_Series.py", line 302, in <module>
obtain_prediction()
File "/home/paul/workspace/workspace/Master/Elec_Price_Prediction/Time_Series.py", line 212, in obtain_prediction
estimator.train(input_fn=train_input_fn, steps=10000)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/estimator/estimator.py", line 302, in train
loss = self._train_model(input_fn, hooks, saving_listeners)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/estimator/estimator.py", line 711, in _train_model
features, labels, model_fn_lib.ModeKeys.TRAIN, self.config)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/estimator/estimator.py", line 694, in _call_model_fn
model_fn_results = self._model_fn(features=features, **kwargs)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/timeseries/python/timeseries/head.py", line 201, in create_estimator_spec
return self._train_ops(features)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/timeseries/python/timeseries/head.py", line 60, in _train_ops
estimator_lib.ModeKeys.TRAIN)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/timeseries/python/timeseries/state_management.py", line 67, in define_loss
return model.define_loss(features, mode)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/timeseries/python/timeseries/model.py", line 196, in define_loss
return self.get_batch_loss(features=features, mode=mode, state=start_state)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/timeseries/python/timeseries/model.py", line 509, in get_batch_loss
features, mode, state)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/timeseries/python/timeseries/model.py", line 609, in per_step_batch_loss
outputs=["loss"] + self._train_output_names)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/timeseries/python/timeseries/model.py", line 775, in _state_update_loop
loop_vars=initial_loop_arguments)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 2816, in while_loop
result = loop_context.BuildLoop(cond, body, loop_vars, shape_invariants)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 2640, in BuildLoop
pred, body, original_loop_vars, loop_vars, shape_invariants)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 2590, in _BuildLoop
body_result = body(*packed_vars_for_body)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/timeseries/python/timeseries/model.py", line 726, in _state_update_step
state=state)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/timeseries/python/timeseries/model.py", line 605, in _batch_loss_filtering_step
predictions=predictions)
File "/home/paul/workspace/workspace/Master/Elec_Price_Prediction/Time_Series.py", line 105, in _filtering_step
prediction=tf.cond(pred,lambda:true_t,lambda:false_t)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/util/deprecation.py", line 316, in new_func
return func(*args, **kwargs)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 1844, in cond
p_2, p_1 = switch(pred, pred)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 305, in switch
return gen_control_flow_ops._switch(data, pred, name=name)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/gen_control_flow_ops.py", line 562, in _switch
"Switch", data=data, pred=pred, name=name)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/op_def_library.py", line 787, in _apply_op_helper
op_def=op_def)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 2958, in create_op
set_shapes_for_outputs(ret)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 2209, in set_shapes_for_outputs
shapes = shape_func(op)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 2159, in call_with_requiring
return call_cpp_shape_fn(op, require_shape_fn=True)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/common_shapes.py", line 627, in call_cpp_shape_fn
require_shape_fn)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/common_shapes.py", line 691, in _call_cpp_shape_fn_impl
raise ValueError(err.message)
ValueError: Shape must be rank 0 but is rank 1 for 'head/model/while/state_update_step/cond/Switch' (op: 'Switch') with input shapes: [12], [12].
My question would be why this is impossible and how to work around it. I tried checking if pred and true_fn as well as false_fn have the same shape and they do.

Answer 1

I prefer tf.where. How about using tf.where?