Merge branch 'splitv' into 'master'

feat: support splitV op for tensorflow

See merge request deep-computing/mace!1284

tools/device.py

@@ -220,8 +220,8 @@ class DeviceWrapper:
                     "MACE_LOG_TENSOR_RANGE=%d" % (1 if quantize_stat else 0),
                     "%s/%s" % (target_dir, target_name),
                     "--model_name=%s" % model_tag,
-                    "--input_node='%s'" % ",".join(input_nodes),
-                    "--output_node='%s'" % ",".join(output_nodes),
+                    "--input_node=%s" % ",".join(input_nodes),
+                    "--output_node=%s" % ",".join(output_nodes),
                     "--input_shape=%s" % ":".join(input_shapes),
                     "--output_shape=%s" % ":".join(output_shapes),
                     "--input_data_format=%s" % ",".join(input_data_formats),

tools/python/transform/base_converter.py

@@ -236,6 +236,7 @@ class MaceKeyword(object):
     mace_end_axis_str = 'end_axis'
     mace_num_axes_str = 'num_axes'
     mace_num_split_str = 'num_split'
+    mace_size_splits_str = 'size_splits'
     mace_keepdims_str = 'keepdims'
     mace_shape_str = 'shape'
     mace_winograd_filter_transformed = 'is_filter_transformed'
@@ -548,6 +549,7 @@ class ConverterOption(object):
                 # Model structure related transformation
                 TransformerRule.REMOVE_USELESS_OP,
                 TransformerRule.TRANSFORM_FAKE_QUANTIZE,
+                TransformerRule.REMOVE_USELESS_OP,
                 TransformerRule.TRANSFORM_GLOBAL_POOLING,
                 TransformerRule.TRANSFORM_LSTMCELL_ZEROSTATE,
                 TransformerRule.TRANSFORM_BASIC_LSTMCELL,

tools/python/transform/megengine_converter.py

@@ -166,7 +166,7 @@ class MegengineConverter(base_converter.ConverterInterface):
                 if "," in op.input[i]:
                     op_name = op.input[i]
                     op_name = op_name.replace(",", "#")
-                    if (op_name in self._option.input_nodes or \
+                    if (op_name in self._option.input_nodes or
                             op_name in self._option.output_nodes):
                         op.input[i] = op_name
             for i in six.moves.range(len(op.output)):
@@ -195,7 +195,8 @@ class MegengineConverter(base_converter.ConverterInterface):
             kernels_arg.name = MaceKeyword.mace_kernel_str
             kernels_arg.ints.extend(kernel)
         if op_def.type in (MaceOp.Conv2D.name, MaceOp.DepthwiseConv2d.name,
-                           MaceOp.Deconv2D.name, MaceOp.DepthwiseDeconv2d.name):
+                           MaceOp.Deconv2D.name,
+                           MaceOp.DepthwiseDeconv2d.name):
             dilation = [params[mge_dilate_h_str], params[mge_dilate_w_str]]
             dilation_arg = op_def.arg.add()
             dilation_arg.name = MaceKeyword.mace_dilations_str
@@ -426,13 +427,14 @@ class MegengineConverter(base_converter.ConverterInterface):
         # check the case of counting include padding
         mode = mge_op.params["mode"]
         if mode == "AVERAGE_COUNT_EXCLUDE_PADDING" or \
-                (mode == "AVERAGE" and mge_op.params["pad_w"] == 0 and \
+                (mode == "AVERAGE" and mge_op.params["pad_w"] == 0 and
                  mge_op.params["pad_h"] == 0):
             pool_type_arg.i = PoolingType.AVG.value
         elif mode == "MAX":
             pool_type_arg.i = PoolingType.MAX.value
         else:
-            mace_check(False, "AVERAGE pooling should not count padding values")
+            mace_check(False,
+                       "AVERAGE pooling should not count padding values")
 
         self.add_stride_pad_kernel_arg(mge_op.params, op)
 

tools/python/transform/tensorflow_converter.py

@@ -116,6 +116,7 @@ TFSupportedOps = [
     'SpaceToBatchND',
     'SpaceToDepth',
     'Split',
+    'SplitV',
     'Sqrt',
     'Square',
     'SquaredDifference',
@@ -279,6 +280,7 @@ class TensorflowConverter(base_converter.ConverterInterface):
             TFOpType.SpaceToBatchND.name: self.convert_space_batch,
             TFOpType.SpaceToDepth.name: self.convert_space_depth,
             TFOpType.Split.name: self.convert_split,
+            TFOpType.SplitV.name: self.convert_splitv,
             TFOpType.Sqrt.name: self.convert_elementwise,
             TFOpType.Squeeze.name: self.convert_squeeze,
             TFOpType.Stack.name: self.convert_stack,
@@ -1057,14 +1059,15 @@ class TensorflowConverter(base_converter.ConverterInterface):
         keep_dims_arg.name = MaceKeyword.mace_keepdims_str
         keep_dims_arg.i = 0
 
-    def convert_split(self, tf_op):
+    def convert_split(self, tf_op, axis_idx=0):
         op = self.convert_general_op(tf_op)
         num_or_size_splits = tf_op.get_attr('num_split')
-        if num_or_size_splits == 1:
+        is_split = (num_or_size_splits > 1)
+        if not is_split:
             op.type = MaceOp.Identity.name
         else:
             op.type = MaceOp.Split.name
-            axis = tf_op.inputs[0].eval().astype(np.int32)
+            axis = tf_op.inputs[axis_idx].eval().astype(np.int32)
             axis = len(op.output_shape[0].dims) + axis if axis < 0 else axis
 
             axis_arg = op.arg.add()
@@ -1074,8 +1077,24 @@ class TensorflowConverter(base_converter.ConverterInterface):
             num_split_arg = op.arg.add()
             num_split_arg.name = MaceKeyword.mace_num_split_str
             num_split_arg.i = num_or_size_splits
-        del op.input[0]
-        self._skip_tensor.add(tf_op.inputs[0].name)
+        del op.input[axis_idx]
+        self._skip_tensor.add(tf_op.inputs[axis_idx].name)
+        return (op, is_split)
+
+    def convert_splitv(self, tf_op):
+        (op, is_split) = self.convert_split(tf_op, 2)
+        if not is_split:
+            return
+        size_splits_arg = op.arg.add()
+        size_splits_arg.name = MaceKeyword.mace_size_splits_str
+        size_splits = tf_op.inputs[1].eval().astype(np.int32)
+        del op.input[1]
+        self._skip_tensor.add(tf_op.inputs[1].name)
+        # todo(luxuhui): support size_splits
+        for size in size_splits:
+            mace_check(size == size_splits[0],
+                       "SplitV Only support even distribution")
+        size_splits_arg.ints.extend(size_splits)
 
     def convert_tile(self, tf_op):
         op = self.convert_general_op(tf_op)

tools/python/validate.py

@@ -318,6 +318,7 @@ def validate_onnx_model(model_file,
                        mace_out_value, value,
                        validation_threshold, log_file)
 
+
 def validate_megengine_model(model_file, input_file,
                              mace_out_file, input_names, input_shapes,
                              input_data_formats, output_names, output_shapes,
@@ -337,7 +338,7 @@ def validate_megengine_model(model_file, input_file,
             util.formatted_file_name(input_file, input_names[i]),
             input_data_types[i])
         input_value = input_value.reshape(input_shapes[i])
-        if (input_data_formats[i] == DataFormat.NHWC and \
+        if (input_data_formats[i] == DataFormat.NHWC and
                 len(input_shapes[i]) == 4):
             input_value = input_value.transpose((0, 3, 1, 2))
         feed_inputs.append(input_value)
@@ -356,10 +357,10 @@ def validate_megengine_model(model_file, input_file,
         output_file_name = \
             util.formatted_file_name(mace_out_file, output_names[i])
         mace_out_value = load_data(output_file_name)
-        if (output_data_formats[i] == DataFormat.NHWC and \
+        if (output_data_formats[i] == DataFormat.NHWC and
                 len(output_shapes[i]) == 4):
-            mace_out_value = \
-                mace_out_value.reshape(output_shapes[i]).transpose((0, 3, 1, 2))
+            mace_out_value = mace_out_value.reshape(
+                output_shapes[i]).transpose((0, 3, 1, 2))
         compare_output(output_names[i], mace_out_value,
                        mge_output_value, validation_threshold, log_file)
 

tools/sh_commands.py

@@ -748,8 +748,8 @@ def validate_model(abi,
                 "--input_file=/mace/%s" % input_file_name,
                 "--mace_out_file=/mace/%s" % output_file_name,
                 "--device_type=%s" % device_type,
-                "--input_node='%s'" % ",".join(input_nodes),
-                "--output_node='%s'" % ",".join(output_nodes),
+                "--input_node=%s" % ",".join(input_nodes),
+                "--output_node=%s" % ",".join(output_nodes),
                 "--input_shape=%s" % ":".join(input_shapes),
                 "--output_shape=%s" % ":".join(output_shapes),
                 "--input_data_format=%s" % ",".join(input_data_formats),

tools/validate.py

@@ -350,7 +350,7 @@ def validate_megengine_model(platform, device_type, model_file, input_file,
             common.formatted_file_name(input_file, input_names[i]),
             input_data_types[i])
         input_value = input_value.reshape(input_shapes[i])
-        if (input_data_formats[i] == common.DataFormat.NHWC and \
+        if (input_data_formats[i] == common.DataFormat.NHWC and
                 len(input_shapes[i]) == 4):
             input_value = input_value.transpose((0, 3, 1, 2))
         feed_inputs.append(input_value)
@@ -369,10 +369,10 @@ def validate_megengine_model(platform, device_type, model_file, input_file,
         output_file_name = \
             common.formatted_file_name(mace_out_file, output_names[i])
         mace_out_value = load_data(output_file_name)
-        if (output_data_formats[i] == common.DataFormat.NHWC and \
+        if (output_data_formats[i] == common.DataFormat.NHWC and
                 len(output_shapes[i]) == 4):
-            mace_out_value = \
-                mace_out_value.reshape(output_shapes[i]).transpose((0, 3, 1, 2))
+            mace_out_value = mace_out_value.reshape(
+                output_shapes[i]).transpose((0, 3, 1, 2))
         compare_output(platform, device_type, output_names[i], mace_out_value,
                        mge_output_value, validation_threshold, log_file)