Add pattern, and remove max-version limit of onnxoptimizer

mindinsight/mindconverter/graph_based_converter/constant.py

@@ -139,7 +139,7 @@ THIRD_PART_VERSION = {
     "torch": (TORCH_MIN_VER,),
     "onnx": (ONNX_MIN_VER,),
     "onnxruntime": (ONNXRUNTIME_MIN_VER,),
-    "onnxoptimizer": (ONNXOPTIMIZER_MIN_VER, ONNXOPTIMIZER_MAX_VER),
+    "onnxoptimizer": (ONNXOPTIMIZER_MIN_VER,),
     "tf2onnx": (TF2ONNX_MIN_VER,)
 }
 

mindinsight/mindconverter/graph_based_converter/framework.py

@@ -26,7 +26,7 @@ from mindinsight.mindconverter.graph_based_converter.common.global_context impor
 from mindinsight.mindconverter.graph_based_converter.common.utils import lib_version_satisfied, onnx_satisfied, \
     save_code_file_and_report, get_framework_type, check_dependency_integrity, get_third_part_lib_validation_error_info
 from mindinsight.mindconverter.graph_based_converter.constant import FrameworkType, \
-    ONNX_MIN_VER, TF2ONNX_MIN_VER, ONNXRUNTIME_MIN_VER, ONNXOPTIMIZER_MIN_VER, ONNXOPTIMIZER_MAX_VER, TORCH_MIN_VER
+    ONNX_MIN_VER, TF2ONNX_MIN_VER, ONNXRUNTIME_MIN_VER, ONNXOPTIMIZER_MIN_VER, TORCH_MIN_VER
 from mindinsight.mindconverter.graph_based_converter.generator import batch_add_nodes
 from mindinsight.mindconverter.graph_based_converter.mapper import ONNXToMindSporeMapper
 from mindinsight.mindconverter.common.log import logger as log, logger_console as log_console
@@ -35,8 +35,6 @@ from mindinsight.mindconverter.common.exceptions import GraphInitError, TreeCrea
     BadParamError
 from mindinsight.mindconverter.graph_based_converter.third_party_graph import GraphFactory
 
-
-
 check_common_dependency_integrity = partial(check_dependency_integrity,
                                             "onnx", "onnxruntime", "onnxoptimizer")
 
@@ -51,7 +49,7 @@ def onnx_lib_version_satisfied():
                             ONNXRUNTIME_MIN_VER, ort.__version__)
 
     if not lib_version_satisfied(getattr(onnx, "__version__"), ONNX_MIN_VER) \
-            or not lib_version_satisfied(getattr(optimizer, "version"), ONNXOPTIMIZER_MIN_VER, ONNXOPTIMIZER_MAX_VER):
+            or not lib_version_satisfied(getattr(optimizer, "version"), ONNXOPTIMIZER_MIN_VER):
         return False
     return True
 

mindinsight/mindconverter/graph_based_converter/sub_graph_searcher/built_in_pattern.py

@@ -175,6 +175,16 @@ def _multi_head_attention_with_einsum():
     ]
 
 
+@register_pattern("Multi-Head-Attention-TF", 2, 1)
+@register_module_name("MultiHeadAttn", 2, 1)
+def _multi_head_attention_tf():
+    return [
+        "MatMul", "Reshape", "Transpose", "MatMul", "Reshape", "Transpose",
+        "MatMul", "Reshape", "Transpose", "MatMul",
+        "Mul", "Add", "Softmax", "MatMul", "Transpose", "Reshape", "MatMul"
+    ]
+
+
 @register_pattern("Layer-Normalization", 1, 1)
 @register_module_name("LayerNorm", 1, 1)
 def _layer_norm():
@@ -183,6 +193,22 @@ def _layer_norm():
     ]
 
 
+@register_pattern("Layer-Normalization-TF", 1, 1)
+@register_module_name("LayerNorm", 1, 1)
+def _layer_norm_tf():
+    return [
+        "ReduceMean", "Sub", "Mul", "ReduceMean", "Add", "Sqrt", "Reciprocal", "Mul", "Mul", "Neg", "Mul", "Add"
+    ]
+
+
+@register_pattern("Feed-Forward-Network-TF", 1, 1)
+@register_module_name("FFN", 1, 1)
+def _ffn_tf():
+    return [
+        "MatMul", "Pow", "Mul", "Add", "Mul", "Tanh", "Add", "Mul", "Mul", "MatMul"
+    ]
+
+
 @register_pattern("Layer-Normalization-with-cast", 1, 1)
 @register_module_name("LayerNorm", 1, 1)
 def _layer_norm_with_cast():