Fixes keras-team#2110 "ValueError: I/O operation on closed file"

This is a workaround for keras-team#2110 where calling `model.fit` with verbose=1
using IPython can intermittently raise "ValueError: I/O operation on
closed file".

This exception appears to be caused by an unknown IO bug with IPython
that is raised when updating the ProgbarLogger progress bar . To
workaround this bug and prevent users from unexpectedly losing their
model:

- The minimum progress bar refresh interval is now 0.1 seconds (up from
0.01 seconds).

- Progress bar updates are now wrapped in try/catch blocks that ignore
ValueError exceptions raised when calling progbar.update

An ideal solution would resolve the IPython at the source, however, this
is an important workaround for users who want to use IPython with
verbose=1.

keras/callbacks.py

@@ -167,7 +167,8 @@ def on_epoch_begin(self, epoch, logs={}):
         if self.verbose:
             print('Epoch %d/%d' % (epoch + 1, self.nb_epoch))
             self.progbar = Progbar(target=self.params['nb_sample'],
-                                   verbose=self.verbose)
+                                   verbose=self.verbose,
+                                   interval=0.1)
         self.seen = 0
 
     def on_batch_begin(self, batch, logs={}):
@@ -185,14 +186,20 @@ def on_batch_end(self, batch, logs={}):
         # skip progbar update for the last batch;
         # will be handled by on_epoch_end
         if self.verbose and self.seen < self.params['nb_sample']:
-            self.progbar.update(self.seen, self.log_values)
+            try:
+                self.progbar.update(self.seen, self.log_values)
+            except ValueError:
+                pass
 
     def on_epoch_end(self, epoch, logs={}):
         for k in self.params['metrics']:
             if k in logs:
                 self.log_values.append((k, logs[k]))
         if self.verbose:
-            self.progbar.update(self.seen, self.log_values, force=True)
+            try:
+                self.progbar.update(self.seen, self.log_values, force=True)
+            except ValueError:
+                pass
 
 
 class History(Callback):