update scale collection and propagation algorithm (#31783) (#31810)

python/paddle/fluid/contrib/slim/quantization/quant2_int8_mkldnn_pass.py

@@ -62,9 +62,8 @@ class Quant2Int8MkldnnPass(object):
         self._ops_to_quantize = _ops_to_quantize
         self._op_ids_to_skip = _op_ids_to_skip if _op_ids_to_skip is not None else set(
             [-1])
-        self._scale_immutable_ops = [
-            'transpose2', 'reshape2', 'pool2d', 'scale'
-        ]
+        self._scale_immutable_ops = ['transpose2', 'reshape2', 'pool2d']
+        self._scale_ops = ['scale']
         self._conv_ops = ['conv2d', 'depthwise_conv2d']
         self._pool_ops = ['pool2d']
         self._mul_ops = ['mul']
@@ -87,8 +86,8 @@ class Quant2Int8MkldnnPass(object):
         self._reset_pass_idx_and_group('int8')
         graph = self._label_skip_quantized_op(graph)
         graph = self._gather_weight_thresholds_from_fake(graph)
-        graph = self._gather_output_scales_from_attr(graph)
         graph = self._gather_input_scales_from_fake(graph)
+        graph = self._gather_output_scales_from_attr(graph)
         graph = self._remove_fake_ops(graph)
         graph = self._dequantize_weights(graph)
         graph = self._optimize_fp32_graph(graph)
@@ -160,12 +159,16 @@ class Quant2Int8MkldnnPass(object):
                     op_node.op()._set_attr("skip_quant", True)
         return graph
 
-    def _gather_input_scales_from_fake(self, graph):
-        def _add_scale_for_vars(var_names, use_unsigned_int, lod_tensor):
-            scales = self._var_quant_scales
-            for var_name in var_names:
+    def _add_scale_for_vars(self, var_names, use_unsigned_int, lod_tensor):
+        """
+        Save quantization scales for variables. Do not overwrite.
+        """
+        scales = self._var_quant_scales
+        for var_name in var_names:
+            if var_name not in scales:
                 scales[var_name] = (use_unsigned_int, lod_tensor)
 
+    def _gather_input_scales_from_fake(self, graph):
         # fake_quantize_dequantize_abs_max doesn't have scale value
         fake_ops = ['fake_quantize_dequantize_moving_average_abs_max']
         fake_ops.extend(self._fake_quantize_types)
@@ -185,8 +188,8 @@ class Quant2Int8MkldnnPass(object):
                 scale[scale == np.Inf] = 0.0
                 lod_tensor = self._convert_scale2tensor(scale)
                 use_unsigned_int = False
-                _add_scale_for_vars([input_name, output_name], use_unsigned_int,
-                                    lod_tensor)
+                self._add_scale_for_vars([input_name, output_name],
+                                         use_unsigned_int, lod_tensor)
 
         return graph
 
@@ -219,8 +222,8 @@ class Quant2Int8MkldnnPass(object):
                 use_unsigned_int = False
                 for output_name in op.op().outputs():
                     for out_var_name in op.op().output(output_name):
-                        self._var_quant_scales[out_var_name] = (
-                            use_unsigned_int, scale_lod_tensor)
+                        self._add_scale_for_vars(
+                            [out_var_name], use_unsigned_int, scale_lod_tensor)
 
         return graph
 
@@ -239,24 +242,21 @@ class Quant2Int8MkldnnPass(object):
                     output_name = op.output("Out")[0]
                     tensor_names = [input_name, output_name]
 
-                    # Scale is not quantized, so if it doesn't have any scales
-                    # to propagate, its tensors won't be added to the waiting list.
-                    if all(name not in self._var_quant_scales for name in tensor_names) \
-                            and op.name() != 'scale':
+                    if all(name not in self._var_quant_scales
+                           for name in tensor_names):
                         waiting_for_scale.update(tensor_names)
                         continue
-
-                    if input_name in self._var_quant_scales:
+                    elif input_name in self._var_quant_scales:
                         self._var_quant_scales[
                             output_name] = self._var_quant_scales[input_name]
                     elif output_name in self._var_quant_scales:
-                        if op.name() == 'scale':
-                            _update_scale_op_in_scale(op, input_name,
-                                                      output_name)
-                        else:
-                            self._var_quant_scales[
-                                input_name] = self._var_quant_scales[
-                                    output_name]
+                        self._var_quant_scales[
+                            input_name] = self._var_quant_scales[output_name]
+                elif op.name() in self._scale_ops:
+                    input_name = op.input("X")[0]
+                    output_name = op.output("Out")[0]
+                    if output_name in self._var_quant_scales:
+                        _update_scale_op_in_scale(op, input_name, output_name)
             return waiting_for_scale
 
         waiting_for_scale = _update_scales(graph)