Use `logger.get_logger_dir()` to access logger directory.

examples/FasterRCNN/train.py

@@ -231,7 +231,7 @@ class EvalCallback(Callback):
     def _eval(self):
         all_results = eval_on_dataflow(self.df, lambda img: detect_one_image(img, self.pred))
         output_file = os.path.join(
-            logger.LOG_DIR, 'outputs{}.json'.format(self.global_step))
+            logger.get_logger_dir(), 'outputs{}.json'.format(self.global_step))
         with open(output_file, 'w') as f:
             json.dump(all_results, f)
         print_evaluation_scores(output_file)

tensorpack/callbacks/graph.py

@@ -134,7 +134,8 @@ class ProcessTensors(Callback):
 class DumpTensors(ProcessTensors):
     """
     Dump some tensors to a file.
-    Every step this callback fetches tensors and write them to a npz file under ``logger.LOG_DIR``.
+    Every step this callback fetches tensors and write them to a npz file
+    under ``logger.get_logger_dir``.
     The dump can be loaded by ``dict(np.load(filename).items())``.
     """
     def __init__(self, names):
@@ -144,7 +145,7 @@ class DumpTensors(ProcessTensors):
         """
         assert isinstance(names, (list, tuple)), names
         self._names = names
-        dir = logger.LOG_DIR
+        dir = logger.get_logger_dir()
 
         def fn(*args):
             dic = {}

tensorpack/callbacks/monitor.py

@@ -193,14 +193,14 @@ class TFEventWriter(TrainingMonitor):
     Write summaries to TensorFlow event file.
     """
     def __new__(cls):
-        if logger.LOG_DIR:
+        if logger.get_logger_dir():
             return super(TFEventWriter, cls).__new__(cls)
         else:
             logger.warn("logger directory was not set. Ignore TFEventWriter.")
             return NoOpMonitor()
 
     def _setup_graph(self):
-        self._writer = tf.summary.FileWriter(logger.LOG_DIR, graph=tf.get_default_graph())
+        self._writer = tf.summary.FileWriter(logger.get_logger_dir(), graph=tf.get_default_graph())
 
     def process_summary(self, summary):
         self._writer.add_summary(summary, self.global_step)
@@ -222,7 +222,7 @@ def TFSummaryWriter(*args, **kwargs):
 
 class JSONWriter(TrainingMonitor):
     """
-    Write all scalar data to a json file under ``logger.LOG_DIR``, grouped by their global step.
+    Write all scalar data to a json file under ``logger.get_logger_dir()``, grouped by their global step.
     This monitor also attemps to recover the epoch number during setup,
     if an existing json file is found at the same place.
     """
@@ -233,14 +233,14 @@ class JSONWriter(TrainingMonitor):
     """
 
     def __new__(cls):
-        if logger.LOG_DIR:
+        if logger.get_logger_dir():
             return super(JSONWriter, cls).__new__(cls)
         else:
             logger.warn("logger directory was not set. Ignore JSONWriter.")
             return NoOpMonitor()
 
     def _before_train(self):
-        self._dir = logger.LOG_DIR
+        self._dir = logger.get_logger_dir()
         self._fname = os.path.join(self._dir, self.FILENAME)
 
         if os.path.isfile(self._fname):

tensorpack/callbacks/param.py

@@ -184,7 +184,7 @@ class HumanHyperParamSetter(HyperParamSetter):
                 If the pair is not found, the param will not be changed.
         """
         super(HumanHyperParamSetter, self).__init__(param)
-        self.file_name = os.path.join(logger.LOG_DIR, file_name)
+        self.file_name = os.path.join(logger.get_logger_dir(), file_name)
         logger.info("Use {} to set hyperparam: '{}'.".format(
             self.file_name, self.param.readable_name))
 

tensorpack/callbacks/prof.py

@@ -105,7 +105,7 @@ class GPUUtilizationTracker(Callback):
 class GraphProfiler(Callback):
     """
     Enable profiling by installing session hooks,
-    and write metadata or tracing files to ``logger.LOG_DIR``.
+    and write metadata or tracing files to ``logger.get_logger_dir()``.
 
     The tracing files can be loaded from ``chrome://tracing``.
     The metadata files can be processed by
@@ -125,7 +125,7 @@ class GraphProfiler(Callback):
             dump_event(bool): Dump to an event processed by FileWriter and
                 will be shown in TensorBoard.
         """
-        self._dir = logger.LOG_DIR
+        self._dir = logger.get_logger_dir()
         self._dump_meta = bool(dump_metadata)
         self._dump_tracing = bool(dump_tracing)
         self._dump_event = bool(dump_event)

tensorpack/callbacks/saver.py

@@ -27,7 +27,7 @@ class ModelSaver(Callback):
         """
         Args:
             max_to_keep, keep_checkpoint_every_n_hours(int): the same as in ``tf.train.Saver``.
-            checkpoint_dir (str): Defaults to ``logger.LOG_DIR``.
+            checkpoint_dir (str): Defaults to ``logger.get_logger_dir()``.
             var_collections (str or list of str): collection of the variables (or list of collections) to save.
         """
         self._max_to_keep = max_to_keep
@@ -43,7 +43,7 @@ class ModelSaver(Callback):
             var_collections = [var_collections]
         self.var_collections = var_collections
         if checkpoint_dir is None:
-            checkpoint_dir = logger.LOG_DIR
+            checkpoint_dir = logger.get_logger_dir()
         assert checkpoint_dir is not None
         if not tf.gfile.IsDirectory(checkpoint_dir):
             tf.gfile.MakeDirs(checkpoint_dir)
@@ -115,7 +115,7 @@ class MinSaver(Callback):
 
         Note:
             It assumes that :class:`ModelSaver` is used with
-            ``checkpoint_dir=logger.LOG_DIR`` (the default). And it will save
+            ``checkpoint_dir=logger.get_logger_dir()`` (the default). And it will save
             the model to that directory as well.
         """
         self.monitor_stat = monitor_stat
@@ -143,13 +143,13 @@ class MinSaver(Callback):
                 self._save()
 
     def _save(self):
-        ckpt = tf.train.get_checkpoint_state(logger.LOG_DIR)
+        ckpt = tf.train.get_checkpoint_state(logger.get_logger_dir())
         if ckpt is None:
             raise RuntimeError(
                 "Cannot find a checkpoint state. Do you forget to use ModelSaver?")
         path = ckpt.model_checkpoint_path
 
-        newname = os.path.join(logger.LOG_DIR,
+        newname = os.path.join(logger.get_logger_dir(),
                                self.filename or
                                ('max-' + self.monitor_stat if self.reverse else 'min-' + self.monitor_stat))
         files_to_copy = tf.gfile.Glob(path + '*')

tensorpack/callbacks/stats.py

@@ -71,7 +71,7 @@ class InjectShell(Callback):
 
 class DumpParamAsImage(Callback):
     """
-    Dump a tensor to image(s) to ``logger.LOG_DIR`` after every epoch.
+    Dump a tensor to image(s) to ``logger.get_logger_dir()`` after every epoch.
 
     Note that it requires the tensor is directly evaluable, i.e. either inputs
     are not its dependency (e.g. the weights of the model), or the inputs are
@@ -93,7 +93,7 @@ class DumpParamAsImage(Callback):
             self.prefix = op_name
         else:
             self.prefix = prefix
-        self.log_dir = logger.LOG_DIR
+        self.log_dir = logger.get_logger_dir()
         self.scale = scale
 
     def _before_train(self):

tensorpack/tfutils/sessinit.py

@@ -270,14 +270,14 @@ def get_model_loader(filename):
 
 def TryResumeTraining():
     """
-    Try loading latest checkpoint from ``logger.LOG_DIR``, only if there is one.
+    Try loading latest checkpoint from ``logger.get_logger_dir()``, only if there is one.
 
     Returns:
         SessInit: either a :class:`JustCurrentSession`, or a :class:`SaverRestore`.
     """
-    if not logger.LOG_DIR:
+    if not logger.get_logger_dir():
         return JustCurrentSession()
-    path = os.path.join(logger.LOG_DIR, 'checkpoint')
+    path = os.path.join(logger.get_logger_dir(), 'checkpoint')
     if not tf.gfile.Exists(path):
         return JustCurrentSession()
     return SaverRestore(path)

tensorpack/utils/logger.py

@@ -11,7 +11,7 @@ from datetime import datetime
 from six.moves import input
 import sys
 
-__all__ = ['set_logger_dir', 'auto_set_dir']
+__all__ = ['set_logger_dir', 'auto_set_dir', 'get_logger_dir']
 
 
 class _MyFormatter(logging.Formatter):
@@ -80,7 +80,8 @@ def set_logger_dir(dirname, action=None):
 
     Args:
         dirname(str): log directory
-        action(str): an action of ("k","b","d","n","q") to be performed. Will ask user by default.
+        action(str): an action of ("k","b","d","n","q") to be performed
+            when the directory exists. Will ask user by default.
     """
     global LOG_DIR, _FILE_HANDLER
     if _FILE_HANDLER:
@@ -128,3 +129,12 @@ def auto_set_dir(action=None, name=None):
     if name:
         auto_dirname += ':%s' % name
     set_logger_dir(auto_dirname, action=action)
+
+
+def get_logger_dir():
+    """
+    Returns:
+        The logger directory, or None if not set.
+        The directory is used for general logging, tensorboard events, checkpoints, etc.
+    """
+    return LOG_DIR