don't need get_model to return output variable

example_alexnet.py

@@ -15,7 +15,7 @@ from tensorpack.models import *
 from tensorpack.utils import *
 from tensorpack.utils.symbolic_functions import *
 from tensorpack.utils.summary import *
-from tensorpack.utils.callback import *
+from tensorpack.callbacks import *
 from tensorpack.dataflow import *
 
 BATCH_SIZE = 10
@@ -25,7 +25,7 @@ CAPACITY = MIN_AFTER_DEQUEUE + 3 * BATCH_SIZE
 def get_model(inputs, is_training):
     # img: 227x227x3
     is_training = bool(is_training)
-    keep_prob = tf.constant(0.5 if is_training else 0.0)
+    keep_prob = tf.constant(0.5 if is_training else 1.0)
 
     image, label = inputs
 
@@ -73,7 +73,7 @@ def get_model(inputs, is_training):
     tf.add_to_collection(MOVING_SUMMARY_VARS_KEY, wd_cost)
 
     add_param_summary('.*/W')   # monitor histogram of all W
-    return [prob, nr_wrong], tf.add_n([wd_cost, cost], name='cost')
+    return tf.add_n([wd_cost, cost], name='cost')
 
 def get_config():
     basename = os.path.basename(__file__)

example_cifar10.py

@@ -83,7 +83,7 @@ def get_model(inputs, is_training):
     tf.add_to_collection(MOVING_SUMMARY_VARS_KEY, wd_cost)
 
     add_param_summary('.*')   # monitor all variables
-    return [prob, nr_wrong], tf.add_n([cost, wd_cost], name='cost')
+    return tf.add_n([cost, wd_cost], name='cost')
 
 def get_config():
     basename = os.path.basename(__file__)

example_mnist.py

@@ -31,12 +31,10 @@ def get_model(inputs, is_training):
             label_var: bx1 integer
         is_training: a python bool variable
     Returns:
-        (outputs, cost)
-        outputs: a list of output variable
-        cost: the cost to minimize. scalar variable
+        the cost to minimize. scalar variable
     """
     is_training = bool(is_training)
-    keep_prob = tf.constant(0.5 if is_training else 0.0)
+    keep_prob = tf.constant(0.5 if is_training else 1.0)
 
     image, label = inputs
     image = tf.expand_dims(image, 3)    # add a single channel
@@ -83,7 +81,7 @@ def get_model(inputs, is_training):
     tf.add_to_collection(MOVING_SUMMARY_VARS_KEY, wd_cost)
 
     add_param_summary('.*/W')   # monitor histogram of all W
-    return [prob, nr_wrong], tf.add_n([wd_cost, cost], name='cost')
+    return tf.add_n([wd_cost, cost], name='cost')
 
 def get_config():
     basename = os.path.basename(__file__)
@@ -126,6 +124,7 @@ def get_config():
         callbacks=Callbacks([
             SummaryWriter(print_tag=['train_cost', 'train_error']),
             PeriodicSaver(),
+            #ValidationCallback(dataset_test, 'test')
             ValidationError(dataset_test, prefix='test'),
         ]),
         session_config=sess_config,

tensorpack/callbacks/group.py

@@ -31,7 +31,7 @@ def create_test_graph():
             ))
         for v in input_vars:
             Gtest.add_to_collection(INPUT_VARS_KEY, v)
-        output_vars, cost = forward_func(input_vars, is_training=False)
+        cost = forward_func(input_vars, is_training=False)
         yield Gtest
 
 @contextmanager

tensorpack/predict.py

@@ -6,6 +6,7 @@
 import tensorflow as tf
 from itertools import count
 import argparse
+from collections import namedtuple
 import numpy as np
 from tqdm import tqdm
 
@@ -43,8 +44,8 @@ class PredictConfig(object):
                 return a tuple of output list as well as the cost to minimize
             output_var_names: a list of names of the output variable to predict, the
                 variables can be any computable tensor in the graph.
-                if None, will predict everything returned by `get_model_func`
-                (all outputs as well as the cost). Predict only specific output
+                if None, will only calculate the cost returned by `get_model_func`.
+                Predict only specific output (instead of the cost)
                 might be faster and might require only some of the input variables.
         """
         def assert_type(v, tp):
@@ -66,14 +67,13 @@ def get_predict_func(config):
         A prediction function that takes a list of inputs value, and return
         one/a list of output values.
         If `output_var_names` is set, then the prediction function will
-        return a list of output values. If not, will return a list of output
-        values and a cost.
+        return a list of output values. If not, will return a cost.
     """
     output_var_names = config.output_var_names
 
     # input/output variables
     input_vars = config.inputs
-    output_vars, cost_var = config.get_model_func(input_vars, is_training=False)
+    cost_var = config.get_model_func(input_vars, is_training=False)
     input_map = config.input_dataset_mapping
     if input_map is None:
         input_map = input_vars
@@ -81,6 +81,8 @@ def get_predict_func(config):
     # check output_var_names against output_vars
     if output_var_names is not None:
         output_vars = [tf.get_default_graph().get_tensor_by_name(n) for n in output_var_names]
+    else:
+        output_vars = []
 
     describe_model()
 
@@ -96,12 +98,13 @@ def get_predict_func(config):
             results = sess.run(output_vars, feed_dict=feed)
             return results
         else:
-            results = sess.run([cost_var] + output_vars, feed_dict=feed)
+            results = sess.run([cost_var], feed_dict=feed)
             cost = results[0]
-            outputs = results[1:]
-            return outputs, cost
+            return cost
     return run_input
 
+PredictResult = namedtuple('PredictResult', ['input', 'output'])
+
 class DatasetPredictor(object):
     def __init__(self, predict_config, dataset, batch=0):
         """
@@ -118,7 +121,7 @@ class DatasetPredictor(object):
         """ a generator to return prediction for each data"""
         with tqdm(total=self.ds.size()) as pbar:
             for dp in self.ds.get_data():
-                yield [dp, self.predict_func(dp)]
+                yield PredictResult(dp, self.predict_func(dp))
                 pbar.update()
 
     def get_all_result(self):

tensorpack/train.py

@@ -36,8 +36,8 @@ class TrainConfig(object):
                 the dataset
             input_queue: the queue used for input. default to a FIFO queue
                 with capacity 5
-            get_model_func: a function taking `inputs` and `is_training` and
-                return a tuple of output list as well as the cost to minimize
+            get_model_func: a function taking `inputs` and `is_training`, and
+                return the cost to minimize
             batched_model_input: boolean. If yes, `get_model_func` expected batched
                 input in training. Otherwise, expect single data point in
                 training, so that you may do pre-processing and batch them
@@ -127,7 +127,7 @@ def start_train(config):
             with tf.device('/gpu:{}'.format(i)), \
                     tf.name_scope('tower{}'.format(i)) as scope:
                 model_inputs = get_model_inputs()
-                output_vars, cost_var = config.get_model_func(model_inputs, is_training=True)
+                cost_var = config.get_model_func(model_inputs, is_training=True)
                 grads.append(
                     config.optimizer.compute_gradients(cost_var))
 
@@ -141,7 +141,7 @@ def start_train(config):
         grads = average_gradients(grads)
     else:
         model_inputs = get_model_inputs()
-        output_vars, cost_var = config.get_model_func(model_inputs, is_training=True)
+        cost_var = config.get_model_func(model_inputs, is_training=True)
         grads = config.optimizer.compute_gradients(cost_var)
     summary_grads(grads)
     avg_maintain_op = summary_moving_average(cost_var)