(fix #685)

examples/Char-RNN/char-rnn.py

@@ -33,7 +33,7 @@ param.seq_len = 50
 param.grad_clip = 5.
 param.vocab_size = None
 param.softmax_temprature = 1
-param.corpus = 'input.txt'
+param.corpus = None
 
 
 class CharRNNData(RNGDataFlow):
@@ -182,6 +182,7 @@ if __name__ == '__main__':
     parser_sample.add_argument('-t', '--temperature', type=float,
                                default=1, help='softmax temperature')
     parser_train = subparsers.add_parser('train', help='train')
+    parser_train.add_argument('--corpus', help='corpus file', default='input.txt')
     args = parser.parse_args()
     if args.gpu:
         os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
@@ -192,6 +193,7 @@ if __name__ == '__main__':
         sample(args.load, args.start, args.num)
         sys.exit()
     else:
+        param.corpus = args.corpus
         config = get_config()
         if args.load:
             config.session_init = SaverRestore(args.load)

tensorpack/callbacks/inference_runner.py

@@ -156,6 +156,7 @@ class InferenceRunner(InferenceRunnerBase):
         for inf in self.infs:
             inf.before_epoch()
 
+        self._input_source.reset_state()
         # iterate over the data, and run the hooked session
         with _inference_context(), \
                 tqdm.tqdm(total=self._size, **get_tqdm_kwargs()) as pbar:
@@ -265,6 +266,7 @@ class DataParallelInferenceRunner(InferenceRunnerBase):
 
         total = self._size
         nr_tower = len(self._gpus)
+        self._input_source.reset_state()
         with _inference_context():
             with tqdm.tqdm(total=total, **get_tqdm_kwargs()) as pbar:
                 while total >= nr_tower:

tensorpack/input_source/input_source.py

@@ -196,25 +196,25 @@ class QueueInput(FeedfreeInput):
 
             self._dequeue_op = self.queue.dequeue(name='dequeue_for_reset')
 
-    def _reset_state(self):
-        if self._started:   # do not try to clear the queue if there is nothing
-            self.thread.pause()     # pause enqueue
-
-            opt = tf.RunOptions()
-            opt.timeout_in_ms = 2000   # 2s
-            sess = tf.get_default_session()
-            # dequeue until empty
-            try:
-                while True:
-                    sess.run(self._dequeue_op, options=opt)
-            except tf.errors.DeadlineExceededError:
-                pass
-
-            # reset dataflow, start thread
-            self.thread.reinitialize_dataflow()
-            self.thread.resume()
-        else:
-            self._started = True
+    def refill_queue(self):
+        """
+        Clear the queue, then call dataflow.get_data() again and fill into the queue.
+        """
+        self.thread.pause()     # pause enqueue
+
+        opt = tf.RunOptions()
+        opt.timeout_in_ms = 2000   # 2s
+        sess = tf.get_default_session()
+        # dequeue until empty
+        try:
+            while True:
+                sess.run(self._dequeue_op, options=opt)
+        except tf.errors.DeadlineExceededError:
+            pass
+
+        # reset dataflow, start thread
+        self.thread.reinitialize_dataflow()
+        self.thread.resume()
 
     def _create_ema_callback(self):
         """

tensorpack/input_source/input_source_base.py

@@ -137,7 +137,7 @@ class InputSource(object):
         Initialize/reinitialize this InputSource.
         Must be called under a default session.
 
-        For training, it will get called by the trainer in `before_train` callbacks.
+        For training, it will get called once by the trainer in `before_train` callbacks.
         For inference, the :class:`InferenceRunner` will call this method each time it is triggered.
         """
         self._reset_state()