onnx2tf 1.25.15__tar.gz → 1.26.1__tar.gz

{onnx2tf-1.25.15/onnx2tf.egg-info → onnx2tf-1.26.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: onnx2tf
-Version: 1.25.15
+Version: 1.26.1
 Summary: Self-Created Tools to convert ONNX files (NCHW) to TensorFlow/TFLite/Keras format (NHWC). The purpose of this tool is to solve the massive Transpose extrapolation problem in onnx-tensorflow (onnx-tf).
 Home-page: https://github.com/PINTO0309/onnx2tf
 Author: Katsuya Hyodo
@@ -314,7 +314,7 @@ Video speed is adjusted approximately 50 times slower than actual speed.
   docker run --rm -it \
   -v `pwd`:/workdir \
   -w /workdir \
-  ghcr.io/pinto0309/onnx2tf:1.25.15
+  ghcr.io/pinto0309/onnx2tf:1.26.1
 
   or
 
@@ -322,7 +322,7 @@ Video speed is adjusted approximately 50 times slower than actual speed.
   docker run --rm -it \
   -v `pwd`:/workdir \
   -w /workdir \
-  docker.io/pinto0309/onnx2tf:1.25.15
+  docker.io/pinto0309/onnx2tf:1.26.1
 
   or
 
@@ -1529,7 +1529,8 @@ usage: onnx2tf
 [-oiqt]
 [-qt {per-channel,per-tensor}]
 [-cind INPUT_NAME NUMPY_FILE_PATH MEAN STD]
-[-ioqd {int8,uint8}]
+[-iqd {int8,uint8,float32}]
+[-oqd {int8,uint8,float32}]
 [-nuo]
 [-nuonag]
 [-b BATCH_SIZE]
@@ -1686,9 +1687,13 @@ optional arguments:
       and {input_op_name}, {numpy_file_path}, {mean}, and {std} must all be entered.
       Otherwise, an error will occur during the -oiqt stage.
 
-  -ioqd {int8,uint8}, --input_output_quant_dtype {int8,uint8}
-    Input and Output dtypes when doing Full INT8 Quantization.
-    "int8"(default) or "uint8"
+  -iqd {int8,uint8,float32}, --input_quant_dtype {int8,uint8,float32}
+    Input dtypes when doing Full INT8 Quantization.
+    "int8"(default) or "uint8" or "float32"
+
+  -oqd {int8,uint8,float32}, --output_quant_dtype {int8,uint8,float32}
+    Output dtypes when doing Full INT8 Quantization.
+    "int8"(default) or "uint8" or "float32"
 
   -nuo, --not_use_onnxsim
     No optimization by onnx-simplifier is performed.
@@ -2008,7 +2013,8 @@ convert(
   output_integer_quantized_tflite: Optional[bool] = False,
   quant_type: Optional[str] = 'per-channel',
   custom_input_op_name_np_data_path: Optional[List] = None,
-  input_output_quant_dtype: Optional[str] = 'int8',
+  input_quant_dtype: Optional[str] = 'int8',
+  output_quant_dtype: Optional[str] = 'int8',
   not_use_onnxsim: Optional[bool] = False,
   not_use_opname_auto_generate: Optional[bool] = False,
   batch_size: Union[int, NoneType] = None,
@@ -2172,9 +2178,13 @@ convert(
         and {input_op_name}, {numpy_file_path}, {mean}, and {std} must all be entered.
         Otherwise, an error will occur during the -oiqt stage.
 
-    input_output_quant_dtype: Optional[str]
-      Input and Output dtypes when doing Full INT8 Quantization.
-      "int8"(default) or "uint8"
+    input_quant_dtype: Optional[str]
+      Input dtypes when doing Full INT8 Quantization.
+      "int8"(default) or "uint8" or "float32"
+
+    output_quant_dtype: Optional[str]
+      Output dtypes when doing Full INT8 Quantization.
+      "int8"(default) or "uint8" or "float32"
 
     not_use_onnxsim: Optional[bool]
       No optimization by onnx-simplifier is performed.
@@ -2600,7 +2610,7 @@ Do not submit an issue that only contains an amount of information that cannot b
   |14|Unsqueeze|1. "param_target": "inputs"<br>`pre_process_transpose_perm`: Transpose is applied to the tensor before the Unsqueeze operation with the perm specified as pre-processing.<br>2. "param_target": "outputs"<br>`post_process_transpose_perm`: Transpose is applied to the tensor after the Unsqueeze operation with the perm specified as post-processing.<br>3. "param_target": "op"<br>`new_shape`: Specifies directly the shape after Unsqueeze processing.<br>{<br>&nbsp;&nbsp;"op_name": "/backbone/backbone.1/Unsqueeze_1",<br>&nbsp;&nbsp;"param_target": "op",<br>&nbsp;&nbsp;"new_shape": [1,15,15,1]<br>}|
   |15|Reshape|1. "param_target": "inputs"<br>`values`: Value of `shape`<br>`pre_process_transpose_perm`: Transpose is applied to the tensor before the Reshape operation with the perm specified as pre-processing.<br>2. "param_target": "outputs"<br>`post_process_transpose_perm`: Transpose is applied to the tensor after the Reshape operation with the perm specified as post-processing.|
   |16|Resize|1. "param_target": "attributes"<br>`coordinate_transformation_mode`: Value of `coordinate_transformation_mode`<br>`extrapolation_value`: Value of `extrapolation_value`<br>`mode`: Value of `mode`<br>2. "param_target": "inputs"<br>`values`: Value of `roi` or `scales` or `sizes`. `scales`=`[scale_h,scale_w]`,`sizes`=`[h,w]`<br>`pre_process_transpose_perm`: Transpose is applied to the tensor before the Resize operation with the perm specified as pre-processing.<br>3. "param_target": "outputs"<br>`post_process_transpose_perm`: Transpose is applied to the tensor after the Resize operation with the perm specified as post-processing.|
-  |17|Slice|`Slice` implements special replacements separately ignore all automatic conversions and generate `tf.strided_slice` directly by specifying all parameters of `tf.strided_slice` directly.<br>https://www.tensorflow.org/api_docs/python/tf/strided_slice<br>See [replace_slice.json](https://github.com/PINTO0309/onnx2tf/blob/main/replace_slice.json) for a sample description.<br>![20221221222956](https://user-images.githubusercontent.com/33194443/208916732-9987a69a-83a7-4a29-8b77-d97b1812d59c.png)<br>1. "param_target": "op"<br>`begin`: Value of `begin`<br>`end`: Value of `end`<br>`strides`: Value of `strides`<br>`begin_mask`: Value of `begin_mask`<br>`end_mask`: Value of `end_mask`<br>`ellipsis_mask`: Value of `ellipsis_mask`<br>`new_axis_mask`: Value of `new_axis_mask`<br>`shrink_axis_mask`: Value of `shrink_axis_mask`<br>{<br>&nbsp;&nbsp;"op_name": "/Slice",<br>&nbsp;&nbsp;"param_target": "op",<br>&nbsp;&nbsp;"begin": [0,0,1,0],<br>&nbsp;&nbsp;"end": [0,0,0,0],<br>&nbsp;&nbsp;"end_mask": 15<br>}|
+  |17|Slice|`Slice` implements special replacements separately ignore all automatic conversions and generate `tf.strided_slice` directly by specifying all parameters of `tf.strided_slice` directly.<br>https://www.tensorflow.org/api_docs/python/tf/strided_slice<br>See [json_samples/replace_slice.json](https://github.com/PINTO0309/onnx2tf/blob/main/json_samples/replace_slice.json) for a sample description.<br>![20221221222956](https://user-images.githubusercontent.com/33194443/208916732-9987a69a-83a7-4a29-8b77-d97b1812d59c.png)<br>1. "param_target": "op"<br>`begin`: Value of `begin`<br>`end`: Value of `end`<br>`strides`: Value of `strides`<br>`begin_mask`: Value of `begin_mask`<br>`end_mask`: Value of `end_mask`<br>`ellipsis_mask`: Value of `ellipsis_mask`<br>`new_axis_mask`: Value of `new_axis_mask`<br>`shrink_axis_mask`: Value of `shrink_axis_mask`<br>{<br>&nbsp;&nbsp;"op_name": "/Slice",<br>&nbsp;&nbsp;"param_target": "op",<br>&nbsp;&nbsp;"begin": [0,0,1,0],<br>&nbsp;&nbsp;"end": [0,0,0,0],<br>&nbsp;&nbsp;"end_mask": 15<br>}|
   |18|Softmax|1. "param_target": "attributes"<br>`axis`: Value of `axis`. The transpositions corresponding to the specified axis are extrapolated before and after `Softmax`.<br>2. "param_target": "inputs"<br>`values`: Value of `tensor`|
   |19|Split|1. "param_target": "inputs"<br>`values`: Value of `split`<br>2. "param_target": "attributes"<br>`axis`: Value of `axis`.<br>`num_outputs`: Value of `num_outputs`.|
   |20|Sub|1. "param_target": "inputs"<br>`values`: Value of `input`<br>`pre_process_transpose_perm`: Transpose is applied to the tensor before the Sub operation with the perm specified as pre-processing.<br>2. "param_target": "outputs"<br>`post_process_transpose_perm`: Transpose is applied to the tensor after the Sub operation with the perm specified as post-processing.|

{onnx2tf-1.25.15 → onnx2tf-1.26.1}/README.md

@@ -299,7 +299,7 @@ Video speed is adjusted approximately 50 times slower than actual speed.
   docker run --rm -it \
   -v `pwd`:/workdir \
   -w /workdir \
-  ghcr.io/pinto0309/onnx2tf:1.25.15
+  ghcr.io/pinto0309/onnx2tf:1.26.1
 
   or
 
@@ -307,7 +307,7 @@ Video speed is adjusted approximately 50 times slower than actual speed.
   docker run --rm -it \
   -v `pwd`:/workdir \
   -w /workdir \
-  docker.io/pinto0309/onnx2tf:1.25.15
+  docker.io/pinto0309/onnx2tf:1.26.1
 
   or
 
@@ -1514,7 +1514,8 @@ usage: onnx2tf
 [-oiqt]
 [-qt {per-channel,per-tensor}]
 [-cind INPUT_NAME NUMPY_FILE_PATH MEAN STD]
-[-ioqd {int8,uint8}]
+[-iqd {int8,uint8,float32}]
+[-oqd {int8,uint8,float32}]
 [-nuo]
 [-nuonag]
 [-b BATCH_SIZE]
@@ -1671,9 +1672,13 @@ optional arguments:
       and {input_op_name}, {numpy_file_path}, {mean}, and {std} must all be entered.
       Otherwise, an error will occur during the -oiqt stage.
 
-  -ioqd {int8,uint8}, --input_output_quant_dtype {int8,uint8}
-    Input and Output dtypes when doing Full INT8 Quantization.
-    "int8"(default) or "uint8"
+  -iqd {int8,uint8,float32}, --input_quant_dtype {int8,uint8,float32}
+    Input dtypes when doing Full INT8 Quantization.
+    "int8"(default) or "uint8" or "float32"
+
+  -oqd {int8,uint8,float32}, --output_quant_dtype {int8,uint8,float32}
+    Output dtypes when doing Full INT8 Quantization.
+    "int8"(default) or "uint8" or "float32"
 
   -nuo, --not_use_onnxsim
     No optimization by onnx-simplifier is performed.
@@ -1993,7 +1998,8 @@ convert(
   output_integer_quantized_tflite: Optional[bool] = False,
   quant_type: Optional[str] = 'per-channel',
   custom_input_op_name_np_data_path: Optional[List] = None,
-  input_output_quant_dtype: Optional[str] = 'int8',
+  input_quant_dtype: Optional[str] = 'int8',
+  output_quant_dtype: Optional[str] = 'int8',
   not_use_onnxsim: Optional[bool] = False,
   not_use_opname_auto_generate: Optional[bool] = False,
   batch_size: Union[int, NoneType] = None,
@@ -2157,9 +2163,13 @@ convert(
         and {input_op_name}, {numpy_file_path}, {mean}, and {std} must all be entered.
         Otherwise, an error will occur during the -oiqt stage.
 
-    input_output_quant_dtype: Optional[str]
-      Input and Output dtypes when doing Full INT8 Quantization.
-      "int8"(default) or "uint8"
+    input_quant_dtype: Optional[str]
+      Input dtypes when doing Full INT8 Quantization.
+      "int8"(default) or "uint8" or "float32"
+
+    output_quant_dtype: Optional[str]
+      Output dtypes when doing Full INT8 Quantization.
+      "int8"(default) or "uint8" or "float32"
 
     not_use_onnxsim: Optional[bool]
       No optimization by onnx-simplifier is performed.
@@ -2585,7 +2595,7 @@ Do not submit an issue that only contains an amount of information that cannot b
   |14|Unsqueeze|1. "param_target": "inputs"<br>`pre_process_transpose_perm`: Transpose is applied to the tensor before the Unsqueeze operation with the perm specified as pre-processing.<br>2. "param_target": "outputs"<br>`post_process_transpose_perm`: Transpose is applied to the tensor after the Unsqueeze operation with the perm specified as post-processing.<br>3. "param_target": "op"<br>`new_shape`: Specifies directly the shape after Unsqueeze processing.<br>{<br>&nbsp;&nbsp;"op_name": "/backbone/backbone.1/Unsqueeze_1",<br>&nbsp;&nbsp;"param_target": "op",<br>&nbsp;&nbsp;"new_shape": [1,15,15,1]<br>}|
   |15|Reshape|1. "param_target": "inputs"<br>`values`: Value of `shape`<br>`pre_process_transpose_perm`: Transpose is applied to the tensor before the Reshape operation with the perm specified as pre-processing.<br>2. "param_target": "outputs"<br>`post_process_transpose_perm`: Transpose is applied to the tensor after the Reshape operation with the perm specified as post-processing.|
   |16|Resize|1. "param_target": "attributes"<br>`coordinate_transformation_mode`: Value of `coordinate_transformation_mode`<br>`extrapolation_value`: Value of `extrapolation_value`<br>`mode`: Value of `mode`<br>2. "param_target": "inputs"<br>`values`: Value of `roi` or `scales` or `sizes`. `scales`=`[scale_h,scale_w]`,`sizes`=`[h,w]`<br>`pre_process_transpose_perm`: Transpose is applied to the tensor before the Resize operation with the perm specified as pre-processing.<br>3. "param_target": "outputs"<br>`post_process_transpose_perm`: Transpose is applied to the tensor after the Resize operation with the perm specified as post-processing.|
-  |17|Slice|`Slice` implements special replacements separately ignore all automatic conversions and generate `tf.strided_slice` directly by specifying all parameters of `tf.strided_slice` directly.<br>https://www.tensorflow.org/api_docs/python/tf/strided_slice<br>See [replace_slice.json](https://github.com/PINTO0309/onnx2tf/blob/main/replace_slice.json) for a sample description.<br>![20221221222956](https://user-images.githubusercontent.com/33194443/208916732-9987a69a-83a7-4a29-8b77-d97b1812d59c.png)<br>1. "param_target": "op"<br>`begin`: Value of `begin`<br>`end`: Value of `end`<br>`strides`: Value of `strides`<br>`begin_mask`: Value of `begin_mask`<br>`end_mask`: Value of `end_mask`<br>`ellipsis_mask`: Value of `ellipsis_mask`<br>`new_axis_mask`: Value of `new_axis_mask`<br>`shrink_axis_mask`: Value of `shrink_axis_mask`<br>{<br>&nbsp;&nbsp;"op_name": "/Slice",<br>&nbsp;&nbsp;"param_target": "op",<br>&nbsp;&nbsp;"begin": [0,0,1,0],<br>&nbsp;&nbsp;"end": [0,0,0,0],<br>&nbsp;&nbsp;"end_mask": 15<br>}|
+  |17|Slice|`Slice` implements special replacements separately ignore all automatic conversions and generate `tf.strided_slice` directly by specifying all parameters of `tf.strided_slice` directly.<br>https://www.tensorflow.org/api_docs/python/tf/strided_slice<br>See [json_samples/replace_slice.json](https://github.com/PINTO0309/onnx2tf/blob/main/json_samples/replace_slice.json) for a sample description.<br>![20221221222956](https://user-images.githubusercontent.com/33194443/208916732-9987a69a-83a7-4a29-8b77-d97b1812d59c.png)<br>1. "param_target": "op"<br>`begin`: Value of `begin`<br>`end`: Value of `end`<br>`strides`: Value of `strides`<br>`begin_mask`: Value of `begin_mask`<br>`end_mask`: Value of `end_mask`<br>`ellipsis_mask`: Value of `ellipsis_mask`<br>`new_axis_mask`: Value of `new_axis_mask`<br>`shrink_axis_mask`: Value of `shrink_axis_mask`<br>{<br>&nbsp;&nbsp;"op_name": "/Slice",<br>&nbsp;&nbsp;"param_target": "op",<br>&nbsp;&nbsp;"begin": [0,0,1,0],<br>&nbsp;&nbsp;"end": [0,0,0,0],<br>&nbsp;&nbsp;"end_mask": 15<br>}|
   |18|Softmax|1. "param_target": "attributes"<br>`axis`: Value of `axis`. The transpositions corresponding to the specified axis are extrapolated before and after `Softmax`.<br>2. "param_target": "inputs"<br>`values`: Value of `tensor`|
   |19|Split|1. "param_target": "inputs"<br>`values`: Value of `split`<br>2. "param_target": "attributes"<br>`axis`: Value of `axis`.<br>`num_outputs`: Value of `num_outputs`.|
   |20|Sub|1. "param_target": "inputs"<br>`values`: Value of `input`<br>`pre_process_transpose_perm`: Transpose is applied to the tensor before the Sub operation with the perm specified as pre-processing.<br>2. "param_target": "outputs"<br>`post_process_transpose_perm`: Transpose is applied to the tensor after the Sub operation with the perm specified as post-processing.|

{onnx2tf-1.25.15 → onnx2tf-1.26.1}/onnx2tf/__init__.py

@@ -1,3 +1,3 @@
 from onnx2tf.onnx2tf import convert, main
 
-__version__ = '1.25.15'
+__version__ = '1.26.1'

{onnx2tf-1.25.15 → onnx2tf-1.26.1}/onnx2tf/onnx2tf.py

@@ -72,7 +72,8 @@ def convert(
     output_integer_quantized_tflite: Optional[bool] = False,
     quant_type: Optional[str] = 'per-channel',
     custom_input_op_name_np_data_path: Optional[List] = None,
-    input_output_quant_dtype: Optional[str] = 'int8',
+    input_quant_dtype: Optional[str] = 'int8',
+    output_quant_dtype: Optional[str] = 'int8',
     not_use_onnxsim: Optional[bool] = False,
     not_use_opname_auto_generate: Optional[bool] = False,
     batch_size: Optional[int] = None,
@@ -221,9 +222,13 @@ def convert(
             ["input2","input2.npy",[0.3],[0.07]],\n
         ]
 
-    input_output_quant_dtype: Optional[str]
-        Input and Output dtypes when doing Full INT8 Quantization.\n
-        "int8"(default) or "uint8"
+    input_quant_dtype: Optional[str]
+        Input dtypes when doing Full INT8 Quantization.\n
+        "int8"(default) or "uint8" or "float32"
+
+    output_quant_dtype: Optional[str]
+        Output dtypes when doing Full INT8 Quantization.\n
+        "int8"(default) or "uint8" or "float32"
 
     not_use_onnxsim: Optional[bool]
         No optimization by onnx-simplifier is performed.\n
@@ -1693,15 +1698,27 @@ def convert(
                 converter._experimental_disable_per_channel = disable_per_channel
                 converter.unfold_batchmatmul = enable_batchmatmul_unfold
                 converter.representative_dataset = representative_dataset_gen
-                inf_type = None
-                if input_output_quant_dtype == 'int8':
-                    inf_type = tf.int8
-                elif input_output_quant_dtype == 'uint8':
-                    inf_type = tf.uint8
+                inf_type_input = None
+                inf_type_output = None
+                if input_quant_dtype == 'int8':
+                    inf_type_input = tf.int8
+                elif input_quant_dtype == 'uint8':
+                    inf_type_input = tf.uint8
+                elif input_quant_dtype == 'float32':
+                    inf_type_input = tf.float32
+                else:
+                    inf_type_input = tf.int8
+                
+                if output_quant_dtype == 'int8':
+                    inf_type_output = tf.int8
+                elif output_quant_dtype == 'uint8':
+                    inf_type_output = tf.uint8
+                elif output_quant_dtype == 'float32':
+                    inf_type_output = tf.float32
                 else:
-                    inf_type = tf.int8
-                converter.inference_input_type = inf_type
-                converter.inference_output_type = inf_type
+                    inf_type_output = tf.int8
+                converter.inference_input_type = inf_type_input
+                converter.inference_output_type = inf_type_output
                 tflite_model = converter.convert()
                 with open(f'{output_folder_path}/{output_file_name}_full_integer_quant.tflite', 'wb') as w:
                     w.write(tflite_model)
@@ -2128,14 +2145,24 @@ def main():
             'Otherwise, an error will occur during the -oiqt stage.'
     )
     parser.add_argument(
-        '-ioqd',
-        '--input_output_quant_dtype',
+        '-iqd',
+        '--input_quant_dtype',
+        type=str,
+        choices=['int8', 'uint8', 'float32'],
+        default='int8',
+        help=\
+            'Input dtypes when doing Full INT8 Quantization. \n' +
+            '"int8"(default) or "uint8" or "float32"'
+    )
+    parser.add_argument(
+        '-oqd',
+        '--output_quant_dtype',
         type=str,
-        choices=['int8', 'uint8'],
+        choices=['int8', 'uint8', 'float32'],
         default='int8',
         help=\
-            'Input and Output dtypes when doing Full INT8 Quantization. \n' +
-            '"int8"(default) or "uint8"'
+            'Output dtypes when doing Full INT8 Quantization. \n' +
+            '"int8"(default) or "uint8" or "float32"'
     )
     parser.add_argument(
         '-nuo',
@@ -2584,7 +2611,8 @@ def main():
         output_integer_quantized_tflite=args.output_integer_quantized_tflite,
         quant_type=args.quant_type,
         custom_input_op_name_np_data_path=custom_params,
-        input_output_quant_dtype=args.input_output_quant_dtype,
+        input_quant_dtype=args.input_quant_dtype,
+        output_quant_dtype=args.output_quant_dtype,
         not_use_onnxsim=args.not_use_onnxsim,
         not_use_opname_auto_generate=args.not_use_opname_auto_generate,
         batch_size=args.batch_size,

{onnx2tf-1.25.15 → onnx2tf-1.26.1}/onnx2tf/ops/Conv.py

@@ -14,6 +14,7 @@ from tensorflow.python.keras.layers import (
 )
 import onnx_graphsurgeon as gs
 from onnx2tf.utils.common_functions import (
+    get_replacement_parameter,
     get_constant_or_variable,
     get_weights_constant_or_variable,
     get_padding_as_op,
@@ -24,6 +25,7 @@ from onnx2tf.utils.common_functions import (
     transpose_with_flexing_deterrence,
     get_tf_model_inputs,
     onnx_tf_tensor_validation,
+    post_process_transpose,
 )
 from typing import Any, Dict
 from onnx2tf.utils.logging import *
@@ -33,6 +35,7 @@ INF_INDEX_VALUE: int = 4294967296
 
 @print_node_info
 @inverted_operation_enable_disable
+@get_replacement_parameter
 def make_node(
     *,
     graph_node: gs.Node,
@@ -932,6 +935,15 @@ def make_node(
                     dilations,
                 )
 
+    # Post-process transpose
+    tf_layers_dict[graph_node_output.name]['tf_node'] = \
+        post_process_transpose(
+            value_before_transpose=tf_layers_dict[graph_node_output.name]['tf_node'],
+            param_target='outputs',
+            param_name=graph_node.outputs[0].name,
+            **kwargs,
+        )
+
     # Generation of Debug Info
     tf_layers_dict[graph_node_output.name]['tf_node_info'] = \
         make_tf_node_info(

{onnx2tf-1.25.15 → onnx2tf-1.26.1}/onnx2tf/utils/common_functions.py

@@ -5857,11 +5857,18 @@ def correction_process_for_accuracy_errors(
         onnx_output_same_shape_counts = collections.Counter(onnx_output_shape)
         if sum([1 if dim > 1 and cnt > 1 else 0 for dim, cnt in onnx_output_same_shape_counts.items()]) >= 1:
             # Generate dummy op
-            dummy_op = tf_func(
-                input_tensor_1,
-                input_tensor_2,
-            )
-            if dummy_op.shape != tf.TensorShape(None):
+            dummy_op = None
+            tensor_2_candidate_for_transpositions = list(itertools.permutations(range(len(input_tensor_2.shape))))
+            for tensor_2_candidate_for_transposition in tensor_2_candidate_for_transpositions:
+                try:
+                    dummy_op = tf_func(
+                        input_tensor_1,
+                        tf.transpose(a=input_tensor_2, perm=tensor_2_candidate_for_transposition),
+                    )
+                    break
+                except Exception as ex:
+                    pass
+            if dummy_op is not None and dummy_op.shape != tf.TensorShape(None):
                 tf_output_shape = [dim if dim is not None else -1 for dim in dummy_op.shape]
                 number_of_dim_other_than_1 = sum([1 if i != 1 else 0 for i in onnx_output_shape])
                 # Processing continues only if there are two or more dimensions other than 1
@@ -5889,7 +5896,7 @@ def correction_process_for_accuracy_errors(
                             tensor_1_candidate_for_transpositions = \
                                 obtaining_an_inverted_pattern_for_brute_force_validation(tensor_shape=validation_data_1.shape)
                             tensor_2_candidate_for_transpositions = \
-                                obtaining_an_inverted_pattern_for_brute_force_validation(tensor_shape=validation_data_2.shape)
+                                list(itertools.permutations(range(len(validation_data_2.shape))))
                             for tensor_1_candidate_for_transposition in tensor_1_candidate_for_transpositions:
                                 for tensor_2_candidate_for_transposition in tensor_2_candidate_for_transpositions:
                                     try:

{onnx2tf-1.25.15 → onnx2tf-1.26.1/onnx2tf.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: onnx2tf
-Version: 1.25.15
+Version: 1.26.1
 Summary: Self-Created Tools to convert ONNX files (NCHW) to TensorFlow/TFLite/Keras format (NHWC). The purpose of this tool is to solve the massive Transpose extrapolation problem in onnx-tensorflow (onnx-tf).
 Home-page: https://github.com/PINTO0309/onnx2tf
 Author: Katsuya Hyodo
@@ -314,7 +314,7 @@ Video speed is adjusted approximately 50 times slower than actual speed.
   docker run --rm -it \
   -v `pwd`:/workdir \
   -w /workdir \
-  ghcr.io/pinto0309/onnx2tf:1.25.15
+  ghcr.io/pinto0309/onnx2tf:1.26.1
 
   or
 
@@ -322,7 +322,7 @@ Video speed is adjusted approximately 50 times slower than actual speed.
   docker run --rm -it \
   -v `pwd`:/workdir \
   -w /workdir \
-  docker.io/pinto0309/onnx2tf:1.25.15
+  docker.io/pinto0309/onnx2tf:1.26.1
 
   or
 
@@ -1529,7 +1529,8 @@ usage: onnx2tf
 [-oiqt]
 [-qt {per-channel,per-tensor}]
 [-cind INPUT_NAME NUMPY_FILE_PATH MEAN STD]
-[-ioqd {int8,uint8}]
+[-iqd {int8,uint8,float32}]
+[-oqd {int8,uint8,float32}]
 [-nuo]
 [-nuonag]
 [-b BATCH_SIZE]
@@ -1686,9 +1687,13 @@ optional arguments:
       and {input_op_name}, {numpy_file_path}, {mean}, and {std} must all be entered.
       Otherwise, an error will occur during the -oiqt stage.
 
-  -ioqd {int8,uint8}, --input_output_quant_dtype {int8,uint8}
-    Input and Output dtypes when doing Full INT8 Quantization.
-    "int8"(default) or "uint8"
+  -iqd {int8,uint8,float32}, --input_quant_dtype {int8,uint8,float32}
+    Input dtypes when doing Full INT8 Quantization.
+    "int8"(default) or "uint8" or "float32"
+
+  -oqd {int8,uint8,float32}, --output_quant_dtype {int8,uint8,float32}
+    Output dtypes when doing Full INT8 Quantization.
+    "int8"(default) or "uint8" or "float32"
 
   -nuo, --not_use_onnxsim
     No optimization by onnx-simplifier is performed.
@@ -2008,7 +2013,8 @@ convert(
   output_integer_quantized_tflite: Optional[bool] = False,
   quant_type: Optional[str] = 'per-channel',
   custom_input_op_name_np_data_path: Optional[List] = None,
-  input_output_quant_dtype: Optional[str] = 'int8',
+  input_quant_dtype: Optional[str] = 'int8',
+  output_quant_dtype: Optional[str] = 'int8',
   not_use_onnxsim: Optional[bool] = False,
   not_use_opname_auto_generate: Optional[bool] = False,
   batch_size: Union[int, NoneType] = None,
@@ -2172,9 +2178,13 @@ convert(
         and {input_op_name}, {numpy_file_path}, {mean}, and {std} must all be entered.
         Otherwise, an error will occur during the -oiqt stage.
 
-    input_output_quant_dtype: Optional[str]
-      Input and Output dtypes when doing Full INT8 Quantization.
-      "int8"(default) or "uint8"
+    input_quant_dtype: Optional[str]
+      Input dtypes when doing Full INT8 Quantization.
+      "int8"(default) or "uint8" or "float32"
+
+    output_quant_dtype: Optional[str]
+      Output dtypes when doing Full INT8 Quantization.
+      "int8"(default) or "uint8" or "float32"
 
     not_use_onnxsim: Optional[bool]
       No optimization by onnx-simplifier is performed.
@@ -2600,7 +2610,7 @@ Do not submit an issue that only contains an amount of information that cannot b
   |14|Unsqueeze|1. "param_target": "inputs"<br>`pre_process_transpose_perm`: Transpose is applied to the tensor before the Unsqueeze operation with the perm specified as pre-processing.<br>2. "param_target": "outputs"<br>`post_process_transpose_perm`: Transpose is applied to the tensor after the Unsqueeze operation with the perm specified as post-processing.<br>3. "param_target": "op"<br>`new_shape`: Specifies directly the shape after Unsqueeze processing.<br>{<br>&nbsp;&nbsp;"op_name": "/backbone/backbone.1/Unsqueeze_1",<br>&nbsp;&nbsp;"param_target": "op",<br>&nbsp;&nbsp;"new_shape": [1,15,15,1]<br>}|
   |15|Reshape|1. "param_target": "inputs"<br>`values`: Value of `shape`<br>`pre_process_transpose_perm`: Transpose is applied to the tensor before the Reshape operation with the perm specified as pre-processing.<br>2. "param_target": "outputs"<br>`post_process_transpose_perm`: Transpose is applied to the tensor after the Reshape operation with the perm specified as post-processing.|
   |16|Resize|1. "param_target": "attributes"<br>`coordinate_transformation_mode`: Value of `coordinate_transformation_mode`<br>`extrapolation_value`: Value of `extrapolation_value`<br>`mode`: Value of `mode`<br>2. "param_target": "inputs"<br>`values`: Value of `roi` or `scales` or `sizes`. `scales`=`[scale_h,scale_w]`,`sizes`=`[h,w]`<br>`pre_process_transpose_perm`: Transpose is applied to the tensor before the Resize operation with the perm specified as pre-processing.<br>3. "param_target": "outputs"<br>`post_process_transpose_perm`: Transpose is applied to the tensor after the Resize operation with the perm specified as post-processing.|
-  |17|Slice|`Slice` implements special replacements separately ignore all automatic conversions and generate `tf.strided_slice` directly by specifying all parameters of `tf.strided_slice` directly.<br>https://www.tensorflow.org/api_docs/python/tf/strided_slice<br>See [replace_slice.json](https://github.com/PINTO0309/onnx2tf/blob/main/replace_slice.json) for a sample description.<br>![20221221222956](https://user-images.githubusercontent.com/33194443/208916732-9987a69a-83a7-4a29-8b77-d97b1812d59c.png)<br>1. "param_target": "op"<br>`begin`: Value of `begin`<br>`end`: Value of `end`<br>`strides`: Value of `strides`<br>`begin_mask`: Value of `begin_mask`<br>`end_mask`: Value of `end_mask`<br>`ellipsis_mask`: Value of `ellipsis_mask`<br>`new_axis_mask`: Value of `new_axis_mask`<br>`shrink_axis_mask`: Value of `shrink_axis_mask`<br>{<br>&nbsp;&nbsp;"op_name": "/Slice",<br>&nbsp;&nbsp;"param_target": "op",<br>&nbsp;&nbsp;"begin": [0,0,1,0],<br>&nbsp;&nbsp;"end": [0,0,0,0],<br>&nbsp;&nbsp;"end_mask": 15<br>}|
+  |17|Slice|`Slice` implements special replacements separately ignore all automatic conversions and generate `tf.strided_slice` directly by specifying all parameters of `tf.strided_slice` directly.<br>https://www.tensorflow.org/api_docs/python/tf/strided_slice<br>See [json_samples/replace_slice.json](https://github.com/PINTO0309/onnx2tf/blob/main/json_samples/replace_slice.json) for a sample description.<br>![20221221222956](https://user-images.githubusercontent.com/33194443/208916732-9987a69a-83a7-4a29-8b77-d97b1812d59c.png)<br>1. "param_target": "op"<br>`begin`: Value of `begin`<br>`end`: Value of `end`<br>`strides`: Value of `strides`<br>`begin_mask`: Value of `begin_mask`<br>`end_mask`: Value of `end_mask`<br>`ellipsis_mask`: Value of `ellipsis_mask`<br>`new_axis_mask`: Value of `new_axis_mask`<br>`shrink_axis_mask`: Value of `shrink_axis_mask`<br>{<br>&nbsp;&nbsp;"op_name": "/Slice",<br>&nbsp;&nbsp;"param_target": "op",<br>&nbsp;&nbsp;"begin": [0,0,1,0],<br>&nbsp;&nbsp;"end": [0,0,0,0],<br>&nbsp;&nbsp;"end_mask": 15<br>}|
   |18|Softmax|1. "param_target": "attributes"<br>`axis`: Value of `axis`. The transpositions corresponding to the specified axis are extrapolated before and after `Softmax`.<br>2. "param_target": "inputs"<br>`values`: Value of `tensor`|
   |19|Split|1. "param_target": "inputs"<br>`values`: Value of `split`<br>2. "param_target": "attributes"<br>`axis`: Value of `axis`.<br>`num_outputs`: Value of `num_outputs`.|
   |20|Sub|1. "param_target": "inputs"<br>`values`: Value of `input`<br>`pre_process_transpose_perm`: Transpose is applied to the tensor before the Sub operation with the perm specified as pre-processing.<br>2. "param_target": "outputs"<br>`post_process_transpose_perm`: Transpose is applied to the tensor after the Sub operation with the perm specified as post-processing.|