Revert "account for batch padding when updating metrics (#7949)" (#8…

…032)

* Revert "Update executor_group.py (#8003)"

This reverts commit 4aaefa0.

* Revert "account for batch padding when updating metrics (#7949)"

This reverts commit 5db5da9.

python/mxnet/module/executor_group.py

@@ -96,18 +96,6 @@ def _merge_multi_context(outputs, major_axis):
     return rets
 
 
-def _slice_axis(arr, axis, islice):
-    """Slices array along axis"""
-    if axis == 0:
-        # slicing NDArray along axis 0 can avoid copying
-        return arr[islice]
-    elif axis > 0:
-        # pylint: disable=no-member
-        return nd.slice(arr, axis=axis, begin=islice.start, end=islice.stop)
-        # pylint: enable=no-member
-    return arr
-
-
 class DataParallelExecutorGroup(object):
     """A group of executors that lives on a group of devices.
     This is a helper class used to implement data parallelization. Each mini-batch will
@@ -218,7 +206,6 @@ def __init__(self, symbol, contexts, workload, data_shapes, label_shapes, param_
 
         # initialize some instance variables
         self.batch_size = None
-        self.cur_batch_pad = None
         self.slices = None
         self.execs = []
         self._default_execs = None
@@ -413,8 +400,6 @@ def forward(self, data_batch, is_train=None):
 
         """
         _load_data(data_batch, self.data_arrays, self.data_layouts)
-        self.cur_batch_pad = getattr(data_batch, 'pad', None)
-
         if is_train is None:
             is_train = self.for_training
 
@@ -572,23 +557,28 @@ def update_metric(self, eval_metric, labels):
             The metric used for evaluation.
         labels : list of NDArray
             Typically comes from `label` of a `DataBatch`.
+        begin : int
+            Starting index of used outputs.
+        end : int or None
+            Ending index of used outputs.
         """
-        pad = self.cur_batch_pad or 0
-        valid_stop = self.batch_size - pad
         for texec, islice in zip(self.execs, self.slices):
             labels_slice = []
-            outputs_slice = []
-            if islice.start >= valid_stop:
-                break
-            if islice.stop > valid_stop:
-                islice = slice(islice.start, valid_stop)
-            oslice = slice(0, islice.stop - islice.start)
-            for label, laxis in zip(labels, self.label_layouts):
-                labels_slice.append(_slice_axis(label, laxis, islice))
-            for output, oaxis in zip(texec.outputs, self.output_layouts):
-                outputs_slice.append(_slice_axis(output, oaxis, oslice))
+            for label, axis in zip(labels, self.label_layouts):
+                if axis == 0:
+                    # slicing NDArray along axis 0 can avoid copying
+                    labels_slice.append(label[islice])
+                elif axis > 0:
+                    # pylint: disable=no-member
+                    label_my_slice = nd.slice_axis(label, axis=axis, begin=islice.start,
+                                                   end=islice.stop).as_in_context(label.context)
+                    # pylint: enable=no-member
+                    labels_slice.append(label_my_slice)
+                else:
+                    labels_slice.append(label)
+
             labels_ = OrderedDict(zip(self.label_names, labels_slice))
-            preds = OrderedDict(zip(self.output_names, outputs_slice))
+            preds = OrderedDict(zip(self.output_names, texec.outputs))
             eval_metric.update_dict(labels_, preds)
 
     def _bind_ith_exec(self, i, data_shapes, label_shapes, shared_group):

tests/python/unittest/test_module.py

@@ -676,40 +676,6 @@ def test_forward_reshape():
     assert mod.get_outputs()[0].shape == (3, 5)
 
 
-def test_forward_update_metric_pad():
-    d = mx.sym.Variable('data')
-    fc = mx.sym.FullyConnected(data=d, num_hidden=2)
-    sym = mx.sym.SoftmaxOutput(data=fc, name='softmax')
-
-    dshape = (10, 2)
-    data = [mx.nd.zeros(dshape)]
-    lshape = (10,)
-    label = [mx.nd.array(([0] * 8) + ([1] * 2))]
-
-    eval_metric = mx.metric.Accuracy()
-    mod = mx.mod.Module(symbol=sym, data_names=['data'],
-                        label_names=['softmax_label'])
-    mod.bind(data_shapes=[('data', dshape)],
-             label_shapes=[('softmax_label', lshape)])
-    mod.init_params(initializer=mx.initializer.Zero())
-    mod.init_optimizer(optimizer_params={'learning_rate': 0.01})
-
-    # Test that accuracy is 0.8 without padding
-    pad = None
-    data_batch = mx.io.DataBatch(data=data, label=label, pad=pad)
-    mod.forward(data_batch)
-    mod.update_metric(eval_metric, data_batch.label)
-    assert eval_metric.get()[1] == 0.8
-
-    # Test that accuracy is 1.0 with padding
-    pad = 2
-    eval_metric.reset()
-    data_batch = mx.io.DataBatch(data=data, label=label, pad=pad)
-    mod.forward(data_batch)
-    mod.update_metric(eval_metric, data_batch.label)
-    assert eval_metric.get()[1] == 1.0
-
-
 if __name__ == '__main__':
     import nose
     nose.runmodule()