ChainInit & verbose download error

examples/ResNet/README.md

@@ -8,3 +8,6 @@ The train error shown here is a moving average of the error rate of each batch i
 The validation error here is computed on test set.
 
 ![cifar10](https://github.com/ppwwyyxx/tensorpack/raw/master/examples/ResNet/cifar10-resnet.png)
+
+Download model:
+[Cifar10 n=18](https://drive.google.com/open?id=0B308TeQzmFDLeHpSaHAxWGV1WDg)

tensorpack/callbacks/inference.py

@@ -178,7 +178,7 @@ class ClassificationError(Inferencer):
         return [self.wrong_var_name]
 
     def _before_inference(self):
-        self.err_stat = Accuracy()
+        self.err_stat = RatioCounter()
 
     def _datapoint(self, dp, outputs):
         batch_size = dp[0].shape[0]   # assume batched input
@@ -186,7 +186,7 @@ class ClassificationError(Inferencer):
         self.err_stat.feed(wrong, batch_size)
 
     def _after_inference(self):
-        self.trainer.write_scalar_summary(self.summary_name, self.err_stat.accuracy)
+        self.trainer.write_scalar_summary(self.summary_name, self.err_stat.ratio)
 
 class BinaryClassificationStats(Inferencer):
     """ Compute precision/recall in binary classification, given the

tensorpack/dataflow/dataset/ilsvrc.py

@@ -50,7 +50,8 @@ class ILSVRCMeta(object):
 
         proto_path = download(CAFFE_PROTO_URL, self.dir)
         ret = os.system('cd {} && protoc caffe.proto --python_out .'.format(self.dir))
-        assert ret == 0, "caffe proto compilation failed!"
+        assert ret == 0, \
+                "caffe proto compilation failed! Did you install protoc?"
 
     def get_image_list(self, name):
         """

tensorpack/tfutils/sessinit.py

@@ -13,7 +13,7 @@ import six
 from ..utils import logger
 
 __all__ = ['SessionInit', 'NewSession', 'SaverRestore',
-           'ParamRestore',
+           'ParamRestore', 'ChainInit',
            'JustCurrentSession',
            'dump_session_params']
 
@@ -65,7 +65,6 @@ class SaverRestore(SessionInit):
     def _init(self, sess):
         logger.info(
             "Restoring checkpoint from {}.".format(self.path))
-        sess.run(tf.initialize_all_variables())
         chkpt_vars = SaverRestore._read_checkpoint_vars(self.path)
         vars_map = SaverRestore._get_vars_to_restore_multimap(chkpt_vars)
         for dic in SaverRestore._produce_restore_dict(vars_map):
@@ -131,7 +130,6 @@ class ParamRestore(SessionInit):
         self.prms = param_dict
 
     def _init(self, sess):
-        sess.run(tf.initialize_all_variables())
         # allow restore non-trainable variables
         variables = tf.get_collection(tf.GraphKeys.VARIABLES)
         var_dict = dict([v.name, v] for v in variables)
@@ -152,6 +150,21 @@ class ParamRestore(SessionInit):
                 value = value.reshape(varshape)
             sess.run(var.assign(value))
 
+def ChainInit(SessionInit):
+    """ Init a session by a list of SessionInit instance."""
+    def __init__(self, sess_inits, new_session=True):
+        """
+        :params sess_inits: list of `SessionInit` instances.
+        :params new_session: add a `NewSession()` and the beginning, if not there
+        """
+        if new_session and not isinstance(sess_inits[0], NewSession):
+            sess_inits.insert(0, NewSession())
+        self.inits = sess_inits
+
+    def _init(self, sess):
+        for i in self.inits:
+            i.init(sess)
+
 def dump_session_params(path):
     """ Dump value of all trainable variables to a dict and save to `path` as
     npy format, loadable by ParamRestore

tensorpack/train/base.py

@@ -105,6 +105,7 @@ class Trainer(object):
         get_global_step_var()   # ensure there is such var, before finalizing the graph
         callbacks = self.config.callbacks
         callbacks.setup_graph(self)
+        self.sess.run(tf.initialize_all_variables())
         self.config.session_init.init(self.sess)
         tf.get_default_graph().finalize()
         self._start_concurrency()

tensorpack/train/config.py

@@ -47,7 +47,7 @@ class TrainConfig(object):
 
         self.session_config = kwargs.pop('session_config', get_default_sess_config())
         assert_type(self.session_config, tf.ConfigProto)
-        self.session_init = kwargs.pop('session_init', NewSession())
+        self.session_init = kwargs.pop('session_init', JustCurrentSession())
         assert_type(self.session_init, SessionInit)
         self.step_per_epoch = int(kwargs.pop('step_per_epoch'))
         self.starting_epoch = int(kwargs.pop('starting_epoch', 1))

tensorpack/utils/fs.py

@@ -27,10 +27,16 @@ def download(url, dir):
         sys.stdout.write('\r>> Downloading %s %.1f%%' %
                          (fname, float(count * block_size) / float(total_size) * 100.0))
         sys.stdout.flush()
-    fpath, _ = urllib.request.urlretrieve(url, fpath, reporthook=_progress)
-    statinfo = os.stat(fpath)
+    try:
+        fpath, _ = urllib.request.urlretrieve(url, fpath, reporthook=_progress)
+        statinfo = os.stat(fpath)
+        size = statinfo.st_size
+    except:
+        logger.error("Failed to download {}".format(url))
+        raise
+    assert size > 0, "Download an empty file!"
     sys.stdout.write('\n')
-    print('Succesfully downloaded ' + fname + " " + str(statinfo.st_size) + ' bytes.')
+    print('Succesfully downloaded ' + fname + " " + str(size) + ' bytes.')
     return fpath
 
 if __name__ == '__main__':

tensorpack/utils/stat.py

@@ -6,6 +6,7 @@ import numpy as np
 __all__ = ['StatCounter', 'Accuracy', 'BinaryStatistics', 'RatioCounter']
 
 class StatCounter(object):
+    """ A simple counter"""
     def __init__(self):
         self.reset()
 
@@ -35,6 +36,7 @@ class StatCounter(object):
         return max(self._values)
 
 class RatioCounter(object):
+    """ A counter to count ratio of something"""
     def __init__(self):
         self.reset()
 
@@ -57,6 +59,7 @@ class RatioCounter(object):
         return self._tot
 
 class Accuracy(RatioCounter):
+    """ A RatioCounter with a fancy name """
     @property
     def accuracy(self):
         return self.ratio