Skip to content

Mace
Mace

慢慢CG / Mace
与 Fork 源项目一致

Fork自 Xiaomi / Mace

通知 1
Star 0
Fork 0
Mace
Mace
未验证 提交 ab7499a5 编写于 作者: R Riccardo 提交者： GitHub
浏览文件
操作

Add support for AddV2 op in TF (#637) 

Co-authored-by: Nriccardo <rgall@sertiscorp.com>
上级 bed88234
隐藏空白更改
内联 并排
Showing with 3 addition and 0 deletion
tools/python/transform/tensorflow_converter.py
浏览文件 @ ab7499a5
...@@ -54,6 +54,7 @@ tf_axis = 'axis' ...@@ -54,6 +54,7 @@ tf_axis = 'axis'
TFSupportedOps = [ TFSupportedOps = [
'Abs', 'Abs',
'Add', 'Add',
'AddV2',
'ArgMax', 'ArgMax',
'AvgPool', 'AvgPool',
'BatchMatMul', 'BatchMatMul',
...@@ -165,6 +166,7 @@ class TensorflowConverter(base_converter.ConverterInterface): ...@@ -165,6 +166,7 @@ class TensorflowConverter(base_converter.ConverterInterface):
eltwise_type = { eltwise_type = {
TFOpType.Add.name: EltwiseType.SUM, TFOpType.Add.name: EltwiseType.SUM,
TFOpType.AddV2.name: EltwiseType.SUM,
TFOpType.Sub.name: EltwiseType.SUB, TFOpType.Sub.name: EltwiseType.SUB,
TFOpType.Mul.name: EltwiseType.PROD, TFOpType.Mul.name: EltwiseType.PROD,
TFOpType.Div.name: EltwiseType.DIV, TFOpType.Div.name: EltwiseType.DIV,
...@@ -210,6 +212,7 @@ class TensorflowConverter(base_converter.ConverterInterface): ...@@ -210,6 +212,7 @@ class TensorflowConverter(base_converter.ConverterInterface):
self._op_converters = { self._op_converters = {
TFOpType.Abs.name: self.convert_elementwise, TFOpType.Abs.name: self.convert_elementwise,
TFOpType.Add.name: self.convert_add, TFOpType.Add.name: self.convert_add,
TFOpType.AddV2.name: self.convert_add,
TFOpType.ArgMax.name: self.convert_argmax, TFOpType.ArgMax.name: self.convert_argmax,
TFOpType.AvgPool.name: self.convert_pooling, TFOpType.AvgPool.name: self.convert_pooling,
TFOpType.BatchMatMul.name: self.convert_matmul, TFOpType.BatchMatMul.name: self.convert_matmul,
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或