don't add scope for new summary module

tensorpack/callbacks/param.py

@@ -258,6 +258,7 @@ class StatMonitorParamSetter(HyperParamSetter):
             if hist_max > hist_first + self.threshold: # large enough
                 return None
         self.last_changed_epoch = self.epoch_num
-        logger.info("[StatMonitorParamSetter] Triggered, history: " + ','.join(hist))
+        logger.info("[StatMonitorParamSetter] Triggered, history: " +
+                ','.join(map(str, hist)))
         return self.value_func(self.get_current_value())
 

tensorpack/tfutils/summary.py

@@ -110,14 +110,14 @@ def summary_moving_average(tensors=None):
     """
     if tensors is None:
         tensors = tf.get_collection(MOVING_SUMMARY_VARS_KEY)
-    with tf.name_scope('EMA-summary'):
-        # TODO will produce EMA_summary/tower0/xxx. not elegant
-        with tf.name_scope(None):
-            averager = tf.train.ExponentialMovingAverage(
-                0.99, num_updates=get_global_step_var(), name='EMA')
-        avg_maintain_op = averager.apply(tensors)
-        for idx, c in enumerate(tensors):
-            name = re.sub('tower[p0-9]+/', '', c.op.name)
-            tf.summary.scalar(name, averager.average(c))
-        return avg_maintain_op
+
+    # TODO will produce tower0/xxx. not elegant
+    with tf.name_scope(None):
+        averager = tf.train.ExponentialMovingAverage(
+            0.90, num_updates=get_global_step_var(), name='EMA')
+    avg_maintain_op = averager.apply(tensors)
+    for idx, c in enumerate(tensors):
+        name = re.sub('tower[p0-9]+/', '', c.op.name)
+        tf.summary.scalar(name + '-summary', averager.average(c))
+    return avg_maintain_op
 

tensorpack/utils/stats.py

 # -*- coding: UTF-8 -*-
-# File: stat.py
+# File: stats.py
 # Author: Yuxin Wu <ppwwyyxx@gmail.com>
 import numpy as np
 
-__all__ = ['StatCounter', 'Accuracy', 'BinaryStatistics', 'RatioCounter']
+__all__ = ['StatCounter', 'Accuracy', 'BinaryStatistics', 'RatioCounter',
+        'OnlineMoments']
 
 class StatCounter(object):
     """ A simple counter"""
@@ -116,3 +117,32 @@ class BinaryStatistics(object):
         if self.nr_pos == 0:
             return 0
         return 1 - self.recall
+
+class OnlineMoments(object):
+    def __init__(self):
+        self._mean = None
+        self._var = None
+        self._n = 0
+
+    def feed(self, x):
+        self._n += 1
+        if self._mean is None:
+            self._mean = x
+            self._var = 0
+        else:
+            diff = (x - self._mean)
+            ninv = 1.0 / self._n
+            self._mean += diff * ninv
+            self._var = (self._n-2.0)/(self._n-1) * self._var + diff * diff * ninv
+
+    @property
+    def mean(self):
+        return self._mean
+
+    @property
+    def variance(self):
+        return self._var
+
+    @property
+    def std(self):
+        return np.sqrt(self._var)