rewrite method in callbacks

example_mnist.py

@@ -85,6 +85,7 @@ def get_config():
     dataset_train = BatchData(dataset.Mnist('train'), 128)
     dataset_test = BatchData(dataset.Mnist('test'), 256, remainder=True)
     step_per_epoch = dataset_train.size()
+    #step_per_epoch = 3
     #dataset_test = FixedSizeData(dataset_test, 20)
 
     sess_config = get_default_sess_config()

tensorpack/callbacks/base.py

@@ -24,6 +24,7 @@ class Callback(object):
     def before_train(self):
         self.graph = tf.get_default_graph()
         self.sess = tf.get_default_session()
+        self.epoch_num = 0
         self._before_train()
 
     def _before_train(self):
@@ -46,22 +47,25 @@ class Callback(object):
         """
 
     def trigger_epoch(self):
+        self.epoch_num += 1
+        self.global_step = get_global_step()
+        self._trigger_epoch()
+
+    def _trigger_epoch(self):
         """
         Callback to be triggered after every epoch (full iteration of input dataset)
         """
 
+
 class PeriodicCallback(Callback):
     def __init__(self, period):
-        self.__period = period
-        self.epoch_num = 0
+        self.period = period
 
-    def trigger_epoch(self):
-        self.epoch_num += 1
-        if self.epoch_num % self.__period == 0:
-            self.global_step = get_global_step()
-            self._trigger()
+    def _trigger_epoch(self):
+        if self.epoch_num % self.period == 0:
+            self._trigger_periodic()
 
     @abstractmethod
-    def _trigger(self):
+    def _trigger_periodic(self):
         pass
 

tensorpack/callbacks/common.py

@@ -24,7 +24,7 @@ class PeriodicSaver(PeriodicCallback):
             max_to_keep=self.keep_recent,
             keep_checkpoint_every_n_hours=self.keep_freq)
 
-    def _trigger(self):
+    def _trigger_periodic(self):
         self.saver.save(
             tf.get_default_session(),
             self.path,
@@ -40,13 +40,16 @@ class SummaryWriter(Callback):
             self.log_dir, graph_def=self.sess.graph_def)
         tf.add_to_collection(SUMMARY_WRITER_COLLECTION_KEY, self.writer)
         self.summary_op = tf.merge_all_summaries()
+        self.epoch_num = 0
 
-    def trigger_epoch(self):
+    def _trigger_epoch(self):
+        self.epoch_num += 1
         # check if there is any summary
         if self.summary_op is None:
             return
         summary_str = self.summary_op.eval()
         summary = tf.Summary.FromString(summary_str)
+        printed_tag = set()
         for val in summary.value:
             #print val.tag
             val.tag = re.sub('tower[0-9]*/', '', val.tag)
@@ -54,7 +57,12 @@ class SummaryWriter(Callback):
                 assert val.WhichOneof('value') == 'simple_value', \
                     'Cannot print summary {}: not a simple_value summary!'.format(val.tag)
                 logger.info('{}: {:.4f}'.format(val.tag, val.simple_value))
+                printed_tag.add(val.tag)
         self.writer.add_summary(summary, get_global_step())
+        if self.epoch_num == 1:
+            if len(printed_tag) != len(self.print_tag):
+                logger.warn("Tags to print not found in Summary Writer: {}".format(
+                    ", ".join([k for k in self.print_tag if k not in printed_tag])))
 
     def _after_train(self):
         self.writer.close()

tensorpack/callbacks/dump.py

@@ -30,7 +30,7 @@ class DumpParamAsImage(Callback):
         self.var = self.graph.get_tensor_by_name(self.var_name)
         self.epoch_num = 0
 
-    def trigger_epoch(self):
+    def _trigger_epoch(self):
         self.epoch_num += 1
         val = self.sess.run(self.var)
         if self.func is not None:

tensorpack/callbacks/group.py

@@ -73,7 +73,7 @@ class TrainCallbacks(Callback):
         else:
             raise ValueError("Callbacks must contain a SummaryWriter!")
 
-    def before_train(self):
+    def _before_train(self):
         for cb in self.cbs:
             cb.before_train()
         self.writer = tf.get_collection(SUMMARY_WRITER_COLLECTION_KEY)[0]
@@ -86,7 +86,7 @@ class TrainCallbacks(Callback):
         for cb in self.cbs:
             cb.trigger_step()
 
-    def trigger_epoch(self):
+    def _trigger_epoch(self):
         tm = CallbackTimeLogger()
         for cb in self.cbs:
             s = time.time()
@@ -104,7 +104,7 @@ class TestCallbacks(Callback):
     def __init__(self, callbacks):
         self.cbs = callbacks
 
-    def before_train(self):
+    def _before_train(self):
         self.writer = tf.get_collection(SUMMARY_WRITER_COLLECTION_KEY)[0]
         with create_test_session() as sess:
             self.sess = sess
@@ -119,7 +119,7 @@ class TestCallbacks(Callback):
         for cb in self.cbs:
             cb.after_train()
 
-    def trigger_epoch(self):
+    def _trigger_epoch(self):
         if not self.cbs:
             return
         tm = CallbackTimeLogger()
@@ -157,7 +157,7 @@ class Callbacks(Callback):
         self.train = TrainCallbacks(train_cbs)
         self.test = TestCallbacks(test_cbs)
 
-    def before_train(self):
+    def _before_train(self):
         self.train.before_train()
         self.test.before_train()
 
@@ -169,7 +169,7 @@ class Callbacks(Callback):
         self.train.trigger_step()
         # test callback don't have trigger_step
 
-    def trigger_epoch(self):
+    def _trigger_epoch(self):
         self.train.trigger_epoch()
         # TODO test callbacks can be run async?
         self.test.trigger_epoch()

tensorpack/callbacks/validation_callback.py

@@ -68,7 +68,7 @@ class ValidationCallback(PeriodicCallback):
             '{}_cost'.format(self.prefix), cost_avg), self.global_step)
         logger.info("{}_cost: {:.4f}".format(self.prefix, cost_avg))
 
-    def _trigger(self):
+    def _trigger_periodic(self):
         for dp, outputs in self._run_validation():
             pass
 
@@ -95,7 +95,7 @@ class ValidationError(ValidationCallback):
     def _get_output_vars(self):
         return [self.wrong_var]
 
-    def _trigger(self):
+    def _trigger_periodic(self):
         err_stat = Accuracy()
         for dp, outputs in self._run_validation():
             batch_size = dp[0].shape[0]   # assume batched input