[MO] Fix many spelling and grammar errors (#2543)

model-optimizer/extensions/analysis/tf_yolo.py

@@ -86,7 +86,7 @@ class TensorFlowYOLOV1V2Analysis(AnalyzeAction):
                       "\t--input_model <path_to_model>/<model_name>.pb\n" \
                       "\t--batch 1\n" \
                       "\t--tensorflow_use_custom_operations_config <OPENVINO_INSTALL_DIR>/deployment_tools/model_optimizer/extensions/front/tf/<yolo_config>.json\n" \
-                      "All detailed information about conversion of this model can be fount at\n" \
+                      "All detailed information about conversion of this model can be found at\n" \
                       "https://docs.openvinotoolkit.org/latest/_docs_MO_DG_prepare_model_convert_model_tf_specific_Convert_YOLO_From_Tensorflow.html"
             return {'model_type': {'YOLO': get_YOLO_params_by_flavor(flavor)}}, message
         else:
@@ -113,7 +113,7 @@ class TensorFlowYOLOV3Analysis(AnalyzeAction):
                       "\t--input_model <path_to_model>/yolo_v3.pb\n" \
                       "\t--batch 1\n" \
                       "\t--tensorflow_use_custom_operations_config <OPENVINO_INSTALL_DIR>/deployment_tools/model_optimizer/extensions/front/tf/yolo_v3.json\n" \
-                      "Detailed information about conversion of this model can be fount at\n" \
+                      "Detailed information about conversion of this model can be found at\n" \
                       "https://docs.openvinotoolkit.org/latest/_docs_MO_DG_prepare_model_convert_model_tf_specific_Convert_YOLO_From_Tensorflow.html"
             return {'model_type': {'YOLO': get_YOLO_params_by_flavor(flavor)}}, message
         else:

model-optimizer/extensions/back/ShuffleChannelPatternOptimization.py

@@ -138,7 +138,7 @@ class ShuffleChannelFusion(BackReplacementPattern):
     We are able to perform the fusion if the pattern satisfies the conditions:
     1. Pattern input 4D shape is the same as pattern output 4D shape
     2. First Reshape splits channel dimension (1 axis) into two dimensions
-    3. Transpose permutes only splitted dimensions
+    3. Transpose permutes only split dimensions
     4. Second Reshape pack them back
 
     Fixes original models reshape-ability (Smart reshape)
@@ -208,7 +208,7 @@ class DepthToSpaceFusion(BackReplacementPattern):
     We are able to perform the fusion if the pattern satisfies the conditions:
     1. Pattern has 6D input and 4D output
     2. First Reshape splits channel dimension (1 axis) into three dimensions [new_depth, block_size, block_size]
-    3. Transpose permutes splitted dimensions with spatial ones
+    3. Transpose permutes split dimensions with spatial ones
     4. Second Reshape pack block size together with spatial dimension
 
     Fixes original models reshape-ability (Smart reshape)

model-optimizer/extensions/back/SpecialNodesFinalization.py

@@ -80,7 +80,7 @@ class CreateConstNodesReplacement(BackReplacementPattern):
     def replace_pattern(self, graph: Graph, match: dict):
         """
             Adds layers with type 'Const' that produce blob from 'bin' file. The pass finds data nodes with one output which
-            doesn't have edge with 'bin' attribute (or with two outputs and at least one output havent 'bin' attr)
+            doesn't have edge with 'bin' attribute (or with two outputs and at least one output doesn't have 'bin' attr)
             and generate Const op node before the node and data node before the Const node. The data node before 'Const'
             node is needed because the op node dumps input tensors to bin file.
         """

model-optimizer/extensions/front/kaldi/memory_offset_adjustment.py

@@ -88,7 +88,7 @@ class MemoryOffsetAdjustment(FrontReplacementSubgraph):
                               \      |
                                \     |
                                Concat
-    In Left branch we have MemoryOffset with k > 0 so we wait until kth frame will be calcualted. In right branch
+    In Left branch we have MemoryOffset with k > 0 so we wait until kth frame will be calculated. In right branch
     we have no such offsets. As result we Concat (or use in any calculations with more than 1 input) kth frame from
     left branch and 0th from right branch. So we need to add synchronization before Concat node. it can be done with
     MemoryOffset(k) inserted before Concat.

model-optimizer/extensions/front/mxnet/ssd_anchor_reshape.py

@@ -29,7 +29,7 @@ from mo.ops.const import Const
 class SsdPatternAnchorReshape(FrontReplacementSubgraph):
     """
     Find ssd anchors and setup variants values.
-    Need to provide compatibility wit IE DetectionOutpyt layer.
+    Need to provide compatibility with IE DetectionOutput layer.
     """
     enabled = True
     graph_condition = [lambda graph: graph.graph['fw'] == 'mxnet' and graph.graph['cmd_params'].enable_ssd_gluoncv]

model-optimizer/extensions/front/onnx/upsample_ext.py

@@ -52,7 +52,7 @@ class UpsampleFrontExtractor(FrontExtractorOp):
                 )
             if math.fabs(scales[0] - 1) > 1e-5 or math.fabs(scales[1] - 1) > 1e-5:
                 raise Error(
-                    'Upsampling of batch and feature dimentions is not supported for node {}.',
+                    'Upsampling of batch and feature dimensions is not supported for node {}.',
                     node.name
                 )
             height_scale = scales[2]

model-optimizer/extensions/front/onnx/upsample_ext_test.py

@@ -73,7 +73,7 @@ class UpsampleONNXExtractorTest(BaseExtractorsTestingClass):
     def test_invalid_scales(self):
         inp, ref = self._base_attrs()
         inp['scales'] = [1.5, 1.5, 2.0, 2.0]
-        with self.assertRaisesRegex(Error, '.*Upsampling of batch and feature dimentions is not supported for node.*'):
+        with self.assertRaisesRegex(Error, '.*Upsampling of batch and feature dimensions is not supported for node.*'):
             out = self._extract(inp)
 
     def test_invalid_2D_scales(self):

model-optimizer/extensions/front/tf/FlattenToReshape.py

@@ -37,7 +37,7 @@ def is_value_is_constant(val: np.ndarray, const: [int, float]):
 class FlattenToReshapeableReshape(FrontReplacementSubgraph):
     """
     The TensorFlow implementation of the Flatten operation is not reshape-able because the batch size is hardcoded
-    during te constant propagation. This transform sets the 'dim' attribute for the Reshape to [0, -1].
+    during the constant propagation. This transform sets the 'dim' attribute for the Reshape to [0, -1].
     """
     enabled = True
 

model-optimizer/extensions/middle/BinarizeWeightsM1P1.py

@@ -46,7 +46,7 @@ class BinarizeWeightsM1P1(MiddleReplacementPattern):
         transparent.
 
         #TODO Describe how to apply multiplication at output ports -- this is not specified. In the current definition
-        we can pass through only scalar multiplication, but we already requre passing it channel-wise.
+        we can pass through only scalar multiplication, but we already require passing it channel-wise.
     """
     enabled = True
 

model-optimizer/extensions/middle/BlockLSTMtoLSTMSequence.py

@@ -110,7 +110,7 @@ class BlockLSTMtoLSTMSequence(MiddleReplacementPattern):
         concatenated cell states over the whole time sequence -> last cell state
 
         BlockLSTM
-           || out 1 (concatenated cell states comming out of BlockLSTM)
+           || out 1 (concatenated cell states coming out of BlockLSTM)
            \/  in 1
         ConcatV2
            || (concatenation with initial state or another unused data)
@@ -265,10 +265,10 @@ class BlockLSTMtoLSTMSequence(MiddleReplacementPattern):
                     list_of_concatenated_hidden_states_children_node_ids.append(node.id)
 
         if len(list_of_concatenated_hidden_states_children_node_ids) != 1:
-            return  # not supported placement of patten
+            return  # not supported placement of pattern
         conacenated_child_node_id = list_of_concatenated_hidden_states_children_node_ids[0]
         if conacenated_child_node_id != match['after_mul_op_to_the_rest_of_model'].id:
-            return  # not supported placement of patten
+            return  # not supported placement of pattern
 
         gather_indexes = match['gather_0'].in_node(1).value
         if len(gather_indexes) == 1:

model-optimizer/extensions/middle/ONNXRNNSequenceNormalize.py

@@ -24,7 +24,7 @@ from mo.ops.op import Op
 class ONNXRNNSequenceNormalize(MiddleReplacementPattern):
     """
         Convert blobs and shapes of ONNX-like LSTM, GRU, RNN cells to common form (internal for MO).
-        After this normalization pass passes for spliting bidirectional calls and
+        After this normalization pass passes for splitting bidirectional calls and
         multilayer cells will be applied.
 
         This transformation pass involves weights and shapes processing only:

model-optimizer/extensions/middle/ReluQuantizeFuse.py

@@ -142,7 +142,7 @@ class ClampQuantizeMark(MiddleReplacementPattern):
                 return
             max_value = quantize.in_port(2).data.get_value()
             if max_value is None:
-                log.debug('ReluQuantizeFuse: cannot fuse because FakeQuantize op has dynamic input on the 2st port, '
+                log.debug('ReluQuantizeFuse: cannot fuse because FakeQuantize op has dynamic input on the 2nd port, '
                           'levels=`{}`'.format(quantize.levels))
                 return
             if np.all(min_value >= clamp_min) and np.all(max_value <= clamp_max):

model-optimizer/extensions/middle/TensorIteratorInput.py

@@ -144,7 +144,7 @@ class SmartInputMatcher(MiddleReplacementPattern):
             if shape['kind'] == 'op' and shape['op'] == 'Const':
                 start = 0
                 end = shape.value[0]
-                log.warning("You network cannot be reshaped since shapes of placeholders is a contants."
+                log.warning("Your network cannot be reshaped since shapes of placeholders are constants."
                             "Please, provide non-constant shapes. ")
 
         # Create input node with params

model-optimizer/extensions/middle/permute_tensor_iterator.py

@@ -186,7 +186,7 @@ class TransposeTensorIteratorLSTM(MiddleReplacementPattern):
         assert len(data_output_port) == 1
         data_input_port = data_input_port[0]
         data_output_port = data_output_port[0]
-        # Verify that they are really connected to Transpose layers (guarantied by port numbers of TI, see the pattern)
+        # Verify that they are really connected to Transpose layers (guaranteed by port numbers of TI, see the pattern)
         assert ti.in_edge(0)['external_port_id'] == ti.input_port_map[data_input_port]['external_port_id']
         assert ti.out_edge(0)['external_port_id'] == ti.output_port_map[data_output_port]['external_port_id']
 

model-optimizer/extensions/middle/reverse_tensor_iterator.py

@@ -101,7 +101,7 @@ class ReverseTensorIteratorLSTM(MiddleReplacementPattern):
 
         if not self.is_fusable_reverse_sequence(direct_reverse) or \
                 not self.is_fusable_reverse_sequence(inverse_reverse):
-            # we can not merge ReverseSequence with ot equal sequences
+            # we can not merge ReverseSequence without equal sequences
             return
 
         # Modify stride in TI

model-optimizer/extensions/ops/fakequantize.py

@@ -66,7 +66,7 @@ class FakeQuantize(Op):
         inputs = [node.in_node(i) for i in range(5)]
         x, input_low, input_high, output_low, output_high = inputs
         assert x.has_valid('shape')
-        # TODO Check all input[1..4] shapes are broadcastable to intput[0] shape
+        # TODO Check all inputs[1..4] shapes are broadcastable to inputs[0] shape
         assert all([broadcastable(inputs[i].shape, inputs[0].shape) for i in range(1, 5)]), \
             "Not all shapes from FakeQuantize inputs can be broadcasted to input[0] for node {}".format(
                 node.soft_get('name'))

model-optimizer/extensions/ops/sparse_reshape.py

@@ -61,6 +61,6 @@ class SparseReshape(Op):
         output_indices_shape = np.concatenate((input_indices_shape[0:1], new_shape_shape))
         node.out_port(0).data.set_shape(output_indices_shape)
 
-        # TODO: implement constant value propogation for common case
+        # TODO: implement constant value propagation for common case
         if np.array_equal(input_shape_value, output_shape_value) and input_indices_value is not None:
             node.out_port(0).data.set_value(input_indices_value)

model-optimizer/extensions/ops/sparse_segment_sqrtn.py

@@ -28,7 +28,7 @@ class SparseSegmentSqrtN(Op):
             - [0, required] Data tensor from which rows are selected for the sum divided by sqrt of N (ND),
             - [1, required] Tensor of indices of selected rows from the first input tensor along 0 dimension (1D),
             - [2, required] Tensor of segment IDs to which selected rows belong.
-                            Selected rows beloging to the same segment are summed up. The tensor has the same size as the second input.
+                            Selected rows belonging to the same segment are summed up. The tensor has the same size as the second input.
                             Values must be sorted and can be repeated. (1D).
         
         One output:

model-optimizer/extensions/ops/sparse_segment_sum.py

@@ -28,7 +28,7 @@ class SparseSegmentSum(Op):
             - [0, required] Data tensor from which rows are selected for the sum (ND),
             - [1, required] Tensor of indices of selected rows from the first input tensor along 0 dimension (1D),
             - [2, required] Tensor of segment IDs to which selected rows for the sum belong.
-                            Selected rows beloging to the same segment are summed up. The tensor has the same size as the second input.
+                            Selected rows belonging to the same segment are summed up. The tensor has the same size as the second input.
                             Values must be sorted and can be repeated. (1D).
         
         One output:

model-optimizer/extensions/ops/split.py

@@ -110,10 +110,10 @@ class VariadicSplitBase(Op):
         # value propagation
         input_value = node.in_port(0).data.get_value()
         if input_value is not None:
-            splitted = np.split(input_value, idxs[:-1], axis)
+            split = np.split(input_value, idxs[:-1], axis)
             for i, port in node.out_ports().items():
                 if not port.disconnected():
-                    port.data.set_value(splitted[i])
+                    port.data.set_value(split[i])
 
         if op == 'VariadicSplit':
             PermuteInputs().set_input_permutation(node.in_node(1), node, 'input:0', 'axis')

model-optimizer/extensions/ops/tensor_iterator.py

@@ -213,7 +213,7 @@ class TensorIterator(Op):
             internal_node = node_map[internal_node_id]
 
             if internal_node.soft_get('type') != 'Result':
-                # this output wont get out of the body, but it is still Result and needed on non first iterations of TI
+                # this output won't get out of the body, but it is still Result and needed on non first iterations of TI
                 assert 'from_port' in record
                 out_port = TensorIterator.special_port_to_real_port(internal_node, record['from_port'], 'out')
                 assert out_port in internal_node.out_ports() and not internal_node.out_port(out_port).disconnected()
@@ -300,7 +300,7 @@ class TensorIterator(Op):
     def generate_port_map(node: Node, src_port_map):
         """ Extract port_map attributes from node and node.body attributes.
 
-            It iterates over src_port_map and substitude external_port_id, internal_port_id and
+            It iterates over src_port_map and substitute external_port_id, internal_port_id and
             internal_layer_id by real values queried from node ports and node.body attributes.
         """
         result_list = []

model-optimizer/extensions/ops/unique.py

@@ -38,7 +38,7 @@ class Unique(Op):
                             and sorted in the same order as in the input (1D)
             - [1, optional] tensor of indices for each value of the input
                             in the tensor of unique elements (1D)
-            - [2, optional] tensor with a number of occurences for each unique element
+            - [2, optional] tensor with a number of occurrences for each unique element
                             in the input (1D)
     '''
     op = 'Unique'