More data types support in BatchData (fix #983)

examples/FasterRCNN/model_frcnn.py

@@ -214,7 +214,7 @@ def fastrcnn_predictions(boxes, scores):
         # sorted_selection = tf.contrib.framework.sort(selection, direction='ASCENDING')
         sorted_selection = -tf.nn.top_k(-selection, k=tf.size(selection))[0]
 
-        if get_tf_version_tuple() >= (1, 12):
+        if get_tf_version_tuple() >= (1, 13):
             mask = tf.sparse.SparseTensor(indices=tf.expand_dims(sorted_selection, 1),
                                           values=tf.ones_like(sorted_selection, dtype=tf.bool),
                                           dense_shape=output_shape)

examples/FasterRCNN/train.py

@@ -522,6 +522,7 @@ if __name__ == '__main__':
     MODEL = ResNetFPNModel() if cfg.MODE_FPN else ResNetC4Model()
 
     if args.visualize or args.evaluate or args.predict:
+        assert tf.test.is_gpu_available()
         assert args.load
         finalize_configs(is_training=False)
 

tensorpack/dataflow/common.py

@@ -128,22 +128,26 @@ class BatchData(ProxyDataFlow):
                 result.append(
                     [x[k] for x in data_holder])
             else:
-                dt = data_holder[0][k]
-                if type(dt) in list(six.integer_types) + [bool]:
-                    tp = 'int32'
-                elif type(dt) == float:
-                    tp = 'float32'
+                data = data_holder[0][k]
+                if isinstance(data, six.integer_types):
+                    dtype = 'int32'
+                elif type(data) == bool:
+                    dtype = 'bool'
+                elif type(data) == float:
+                    dtype = 'float32'
+                elif isinstance(data, (six.binary_type, six.text_type)):
+                    dtype = 'str'
                 else:
                     try:
-                        tp = dt.dtype
+                        dtype = data.dtype
                     except AttributeError:
-                        raise TypeError("Unsupported type to batch: {}".format(type(dt)))
+                        raise TypeError("Unsupported type to batch: {}".format(type(data)))
                 try:
                     result.append(
-                        np.asarray([x[k] for x in data_holder], dtype=tp))
+                        np.asarray([x[k] for x in data_holder], dtype=dtype))
                 except Exception as e:  # noqa
                     logger.exception("Cannot batch data. Perhaps they are of inconsistent shape?")
-                    if isinstance(dt, np.ndarray):
+                    if isinstance(data, np.ndarray):
                         s = pprint.pformat([x[k].shape for x in data_holder])
                         logger.error("Shape of all arrays to be batched: " + s)
                     try: