min/max saver

examples/OpenAIGym/README.md

@@ -5,10 +5,9 @@
 2. Download models from [model zoo](https://drive.google.com/open?id=0B9IPQTvr2BBkS0VhX0xmS1c5aFk)
 3. `ENV=NAME_OF_ENV ./run-atari.py --load "$ENV".tfmodel --env "$ENV"`
 
-<!--
-   -Models are available for the following gym environments:
-   -
-   -+ [Breakout-v0](https://gym.openai.com/envs/Breakout-v0)
-	 -->
+Models are available for the following gym environments:
 
-Note that atari game settings in gym is very different from DeepMind papers, therefore the scores are not comparable.
++ [Breakout-v0](https://gym.openai.com/envs/Breakout-v0)
++ [AirRaid-v0](https://gym.openai.com/envs/AirRaid-v0)
+
+Note that atari game settings in gym is more difficult than the settings DeepMind papers, therefore the scores are not comparable.

tensorpack/callbacks/common.py

@@ -10,7 +10,7 @@ from .base import Callback
 from ..utils import *
 from ..tfutils.varmanip import get_savename_from_varname
 
-__all__ = ['ModelSaver']
+__all__ = ['ModelSaver', 'MinSaver', 'MaxSaver']
 
 class ModelSaver(Callback):
     """
@@ -81,15 +81,22 @@ because {} will be saved".format(v.name, var_dict[name].name))
             logger.exception("Exception in ModelSaver.trigger_epoch!")
 
 class MinSaver(Callback):
-    def __init__(self, monitor_stat):
+    def __init__(self, monitor_stat, reverse=True):
         self.monitor_stat = monitor_stat
+        self.reverse = reverse
         self.min = None
 
     def _get_stat(self):
         return self.trainer.stat_holder.get_stat_now(self.monitor_stat)
 
+    def _need_save(self):
+        if self.reverse:
+            return self._get_stat() > self.min
+        else:
+            return self._get_stat() < self.min
+
     def _trigger_epoch(self):
-        if self.min is None or self._get_stat() < self.min:
+        if self.min is None or self._need_save():
             self.min = self._get_stat()
             self._save()
 
@@ -97,10 +104,17 @@ class MinSaver(Callback):
         ckpt = tf.train.get_checkpoint_state(logger.LOG_DIR)
         if ckpt is None:
             raise RuntimeError(
-                "Cannot find a checkpoint state. Do you forget to use ModelSaver before MinSaver?")
+                "Cannot find a checkpoint state. Do you forget to use ModelSaver?")
         path = chpt.model_checkpoint_path
-        newname = os.path.join(logger.LOG_DIR, 'min_' + self.monitor_stat)
+        newname = os.path.join(logger.LOG_DIR,
+                'max_' if self.reverse else 'min_' + self.monitor_stat)
         shutil.copy(path, newname)
-        logger.info("Model with minimum {} saved.".format(self.monitor_stat))
+        logger.info("Model with {} {} saved.".format(
+            'maximum' if self.reverse else 'minimum', self.monitor_stat))
+
+class MaxSaver(MinSaver):
+    def __init__(self, monitor_stat):
+        super(MaxSaver, self).__init__(monitor_stat, True)
+
 
 

tensorpack/utils/logger.py

@@ -67,9 +67,9 @@ def set_logger_dir(dirname, action=None):
     if os.path.isdir(dirname):
         if not action:
             _logger.warn("""\
-Directory {} exists! Please either backup/delete it, or use a new directory.""")
+Directory {} exists! Please either backup/delete it, or use a new directory.""".format(dirname))
             _logger.warn("""\
-If you're resuming from a previous run you can choose to keep it.""".format(dirname))
+If you're resuming from a previous run you can choose to keep it.""")
             _logger.info("Select Action: k (keep) / b (backup) / d (delete) / n (new):")
         while not action:
             action = input().lower().strip()