reducers_axes -> reduction_axes, aggregator_axes=aggregation_axes

nncf/torch/quantization/init_range.py

@@ -178,8 +178,8 @@ def generate_stat_collector_for_range_init_config(
 
         use_per_sample_stats = collector_params.use_per_sample_stats(init_config.init_type == "mixed_min_max")
         collector_kwargs = {
-            "reducers_axes": collector_params.get_reduction_axes(use_per_sample_stats),
-            "aggregators_axes": collector_params.get_aggregation_axes(use_per_sample_stats),
+            "reduction_axes": collector_params.get_reduction_axes(use_per_sample_stats),
+            "aggregation_axes": collector_params.get_aggregation_axes(use_per_sample_stats),
             "scale_shape": scale_shape,
         }
 

nncf/torch/quantization/layers.py

@@ -997,8 +997,6 @@ def get_quantizer_config(self) -> QuantizerConfig:
 
 
 def get_per_channel_scale_shape(input_shape, is_weights, channel_idx: int = None):
-    # TODO: case channel_ids=0, is_weights=True and per_sample_stats=True
-    # leads to dimension error in statistic calculation
     scale_shape = [1 for _ in input_shape]
     if channel_idx is None:
         if is_weights:

nncf/torch/tensor_statistics/collectors.py

@@ -237,17 +237,17 @@ class PTMeanPerChanelReducer(PTReducerMixIn, MeanPerChReducer):
 
 def get_min_max_statistic_collector(
     use_abs_max: bool,
-    reducers_axes: Tuple[int, ...],
-    aggregators_axes: Tuple[int, ...],
+    reduction_axes: Tuple[int, ...],
+    aggregation_axes: Tuple[int, ...],
     scale_shape: Tuple[int, ...],
     num_samples: int,
 ) -> TensorCollector:
     """
     Min max statistic collector builder.
 
     :param use_abs_max: Whether to use abs max reducer or max reducer.
-    :param reducers_axes: Axes to use in reduction functions.
-    :param aggregators_axes: Axes to use in aggregation functions.
+    :param reduction_axes: Axes to use in reduction functions.
+    :param aggregation_axes: Axes to use in aggregation functions.
     :param scale_shape: Target shape for collected statistics.
     :param num_samples: Maximum number of samples to collect.
     :return: Min max statistic collector.
@@ -257,23 +257,23 @@ def get_min_max_statistic_collector(
     aggregator_kwargs = {
         "tensor_processor": PTNNCFCollectorTensorProcessor,
         "num_samples": num_samples,
-        "aggregation_axes": aggregators_axes,
+        "aggregation_axes": aggregation_axes,
     }
-    min_reducer = PTMinReducer(reducers_axes)
+    min_reducer = PTMinReducer(reduction_axes)
     min_aggregator = MinAggregator(**aggregator_kwargs)
     tensor_collector.register_statistic_branch(PTMinMaxTensorStatistic.MIN_STAT, min_reducer, min_aggregator)
 
     max_reducer_cls = PTAbsMaxReducer if use_abs_max else PTMaxReducer
-    max_reducer = max_reducer_cls(reducers_axes)
+    max_reducer = max_reducer_cls(reduction_axes)
     max_aggregator = MaxAggregator(**aggregator_kwargs)
     tensor_collector.register_statistic_branch(PTMinMaxTensorStatistic.MAX_STAT, max_reducer, max_aggregator)
     return tensor_collector
 
 
 def get_mixed_min_max_statistic_collector(
     use_abs_max: bool,
-    reducers_axes: Tuple[int, ...],
-    aggregators_axes: Tuple[int, ...],
+    reduction_axes: Tuple[int, ...],
+    aggregation_axes: Tuple[int, ...],
     scale_shape: Tuple[int, ...],
     use_means_of_mins: bool,
     use_means_of_maxs: bool,
@@ -284,8 +284,8 @@ def get_mixed_min_max_statistic_collector(
     Mixed min max statistic collector builder.
 
     :param use_abs_max: Whether to use abs max reducer or max reducer.
-    :param reducers_axes: Axes to use in reduction functions.
-    :param aggregators_axes: Axes to use in aggregation functions.
+    :param reduction_axes: Axes to use in reduction functions.
+    :param aggregation_axes: Axes to use in aggregation functions.
     :param scale_shape: Target shape for collected statistics.
     :param use_means_of_mins: Whether to use mean or min aggregator for minimum statistic branch.
     :param use_means_of_maxs: Whether to use mean or max aggregator for maximum statistic branch.
@@ -295,20 +295,20 @@ def get_mixed_min_max_statistic_collector(
     :return: Mixed min max statistic collector.
     """
     tensor_collector = TensorCollector(partial(PTMinMaxTensorStatistic, target_shape=scale_shape))
-    min_reducer = PTMinReducer(reducers_axes)
+    min_reducer = PTMinReducer(reduction_axes)
 
     kwargs = {
         "tensor_processor": PTNNCFCollectorTensorProcessor,
         "num_samples": num_samples,
-        "aggregation_axes": aggregators_axes,
+        "aggregation_axes": aggregation_axes,
         "window_size": window_size,
     }
     min_aggregator_cls = MeanAggregator if use_means_of_mins else MinAggregator
     min_aggregator = min_aggregator_cls(**kwargs)
     tensor_collector.register_statistic_branch(PTMinMaxTensorStatistic.MIN_STAT, min_reducer, min_aggregator)
 
     max_reducer_cls = PTAbsMaxReducer if use_abs_max else PTMaxReducer
-    max_reducer = max_reducer_cls(reducers_axes)
+    max_reducer = max_reducer_cls(reduction_axes)
     max_aggregator_cls = MeanAggregator if use_means_of_maxs else MaxAggregator
     max_aggregator = max_aggregator_cls(**kwargs)
     tensor_collector.register_statistic_branch(PTMinMaxTensorStatistic.MAX_STAT, max_reducer, max_aggregator)
@@ -317,17 +317,17 @@ def get_mixed_min_max_statistic_collector(
 
 
 def get_median_mad_statistic_collector(
-    reducers_axes: Tuple[int, ...],
-    aggregators_axes: Tuple[int, ...],
+    reduction_axes: Tuple[int, ...],
+    aggregation_axes: Tuple[int, ...],
     scale_shape: Tuple[int, ...],
     num_samples: int,
     window_size: Optional[int] = None,
 ) -> TensorCollector:
     """
     Median Absolute Deviation statistic collector builder.
 
-    :param reducers_axes: Axes to use in reduction functions.
-    :param aggregators_axes: Axes to use in aggregation functions.
+    :param reduction_axes: Axes to use in reduction functions.
+    :param aggregation_axes: Axes to use in aggregation functions.
     :param scale_shape: Target shape for collected statistics.
     :param num_samples: Maximum number of samples to collect.
     :param window_size: Number of samples from the end of the list of collected samples to aggregate.
@@ -338,8 +338,8 @@ def get_median_mad_statistic_collector(
     return _get_collection_without_reduction(
         MedianAbsoluteDeviationAggregator,
         PTMedianMADTensorStatistic,
-        reducers_axes=reducers_axes,
-        aggregators_axes=aggregators_axes,
+        reduction_axes=reduction_axes,
+        aggregation_axes=aggregation_axes,
         scale_shape=scale_shape,
         num_samples=num_samples,
         window_size=window_size,
@@ -348,8 +348,8 @@ def get_median_mad_statistic_collector(
 
 def get_percentile_tensor_collector(
     percentiles_to_collect: Tuple[int, ...],
-    reducers_axes: Tuple[int, ...],
-    aggregators_axes: Tuple[int, ...],
+    reduction_axes: Tuple[int, ...],
+    aggregation_axes: Tuple[int, ...],
     scale_shape: Tuple[int, ...],
     num_samples: int,
     window_size: Optional[int] = None,
@@ -358,8 +358,8 @@ def get_percentile_tensor_collector(
     Percentile statistic collector builder.
 
     :param percentiles_to_collect: Percetiles to use on aggregation phase.
-    :param reducers_axes: Axes to use in reduction functions.
-    :param aggregators_axes: Axes to use in aggregation functions.
+    :param reduction_axes: Axes to use in reduction functions.
+    :param aggregation_axes: Axes to use in aggregation functions.
     :param scale_shape: Target shape for collected statistics.
     :param num_samples: Maximum number of samples to collect.
     :param window_size: Number of samples from the end of the list of collected samples to aggregate.
@@ -369,8 +369,8 @@ def get_percentile_tensor_collector(
     return _get_collection_without_reduction(
         partial(percentileAggregator, percentiles_to_collect=percentiles_to_collect),
         PTPercentileTensorStatistic,
-        reducers_axes=reducers_axes,
-        aggregators_axes=aggregators_axes,
+        reduction_axes=reduction_axes,
+        aggregation_axes=aggregation_axes,
         scale_shape=scale_shape,
         num_samples=num_samples,
         window_size=window_size,
@@ -380,8 +380,8 @@ def get_percentile_tensor_collector(
 def _get_collection_without_reduction(
     aggregator_cls: TensorAggregatorBase,
     statistic_cls: TensorAggregatorBase,
-    reducers_axes: Tuple[int, ...],
-    aggregators_axes: Tuple[int, ...],
+    reduction_axes: Tuple[int, ...],
+    aggregation_axes: Tuple[int, ...],
     scale_shape: Tuple[int, ...],
     num_samples: int,
     window_size: Optional[int] = None,
@@ -391,8 +391,8 @@ def _get_collection_without_reduction(
 
     :param aggregator_cls: Aggregator class to build the tensor collector.
     :param aggregator_cls: Statistic class to build the tensor collector.
-    :param reducers_axes: Axes to use in reduction functions.
-    :param aggregators_axes: Axes to use in aggregation functions.
+    :param reduction_axes: Axes to use in reduction functions.
+    :param aggregation_axes: Axes to use in aggregation functions.
     :param scale_shape: Target shape for collected statistics.
     :param num_samples: Maximum number of samples to collect.
     :param window_size: Number of samples from the end of the list of collected samples to aggregate.
@@ -401,7 +401,7 @@ def _get_collection_without_reduction(
     """
     tensor_collector = TensorCollector(partial(statistic_cls, target_shape=scale_shape))
     reducer = PTNoopReducer()
-    aggregation_axes = list(set(list(aggregators_axes) + [dim + 1 for dim in reducers_axes]))
+    aggregation_axes = list(set(list(aggregation_axes) + [dim + 1 for dim in reduction_axes]))
     aggregator = aggregator_cls(
         PTNNCFCollectorTensorProcessor,
         aggregation_axes=aggregation_axes,
@@ -417,8 +417,8 @@ def _get_collection_without_reduction(
 
 def get_mean_percentile_statistic_collector(
     percentiles_to_collect: Tuple[int, ...],
-    reducers_axes: Tuple[int, ...],
-    aggregators_axes: Tuple[int, ...],
+    reduction_axes: Tuple[int, ...],
+    aggregation_axes: Tuple[int, ...],
     scale_shape: Tuple[int, ...],
     num_samples: int,
     window_size: Optional[int] = None,
@@ -427,8 +427,8 @@ def get_mean_percentile_statistic_collector(
     Mean percentile statistic collector builder.
 
     :param percentiles_to_collect: Percetiles to use on reduction phase.
-    :param reducers_axes: Axes to use in reduction functions.
-    :param aggregators_axes: Axes to use in aggregation functions.
+    :param reduction_axes: Axes to use in reduction functions.
+    :param aggregation_axes: Axes to use in aggregation functions.
     :param scale_shape: Target shape for collected statistics.
     :param num_samples: Maximum number of samples to collect.
     :param window_size: Number of samples from the end of the list of collected samples to aggregate.
@@ -437,11 +437,11 @@ def get_mean_percentile_statistic_collector(
     """
     tensor_collector = TensorCollector(partial(PTPercentileTensorStatistic, target_shape=scale_shape))
     quantiles_to_collect = np.true_divide(percentiles_to_collect, 100)
-    reducer = PTQuantileReducer(reduction_axes=reducers_axes, quantile=quantiles_to_collect)
+    reducer = PTQuantileReducer(reduction_axes=reduction_axes, quantile=quantiles_to_collect)
     for output_port_id, p in enumerate(percentiles_to_collect):
         aggregator = MeanAggregator(
             PTNNCFCollectorTensorProcessor,
-            aggregation_axes=aggregators_axes,
+            aggregation_axes=aggregation_axes,
             num_samples=num_samples,
             window_size=window_size,
         )

tests/torch/tensor_statistics/test_tensor_statistics.py

@@ -132,8 +132,8 @@ def test_collected_statistics_with_shape_convert(
             collector_obj = collector(
                 scale_shape=scale_shape,
                 use_abs_max=True,
-                reducers_axes=reducer_axes,
-                aggregators_axes=(0,),
+                reduction_axes=reducer_axes,
+                aggregation_axes=(0,),
                 num_samples=None,
             )
             for input_ in TestCollectedStatistics.REF_INPUTS:
@@ -246,8 +246,8 @@ def test_collected_statistics(
 
             collector_obj = collector(
                 scale_shape=reduction_shape,
-                reducers_axes=reducer_axes,
-                aggregators_axes=(0,),
+                reduction_axes=reducer_axes,
+                aggregation_axes=(0,),
                 num_samples=None,
             )
             for input_ in TestCollectedStatistics.REF_INPUTS: