Bug: reorder FOLD_BIASADD method for quantization models.

b297e75b · liuqi · f002c764 · b297e75b
隐藏空白更改
内联并排

Showing with 2 addition and 2 deletion

mace/python/tools/converter_tool/base_converter.py mace/python/tools/converter_tool/base_converter.py +2 -2

未找到文件。
--- a/mace/python/tools/converter_tool/base_converter.py
+++ b/mace/python/tools/converter_tool/base_converter.py
@@ -415,12 +415,12 @@ class ConverterOption(object):
                TransformerRule.FOLD_RESHAPE,
                TransformerRule.TRANSFORM_MATMUL_TO_FC,
                TransformerRule.FOLD_BATCHNORM,
-                TransformerRule.TRANSFORM_ADD_TO_BIASADD,
-                TransformerRule.FOLD_BIASADD,
                TransformerRule.FOLD_CONV_AND_BN,
                TransformerRule.FOLD_DECONV_AND_BN,
                TransformerRule.FOLD_DEPTHWISE_CONV_AND_BN,
+                TransformerRule.TRANSFORM_ADD_TO_BIASADD,
                TransformerRule.REARRANGE_BATCH_TO_SPACE,
+                TransformerRule.FOLD_BIASADD,
                TransformerRule.FLATTEN_ATROUS_CONV,
                TransformerRule.FOLD_ACTIVATION,
                TransformerRule.FOLD_SQRDIFF_MEAN,