提交 ddffa5f6 编写于 作者: C chengduozh

fix merge lod_tensor bug

test=release/1.1
上级 2cbf8368
......@@ -418,7 +418,7 @@ void LoDTensor::MergeLoDTensor(
PADDLE_ENFORCE_EQ(new_lod.size(), lod.size());
for (size_t j = 0; j < lod.size(); ++j) {
auto &sub_lod = new_lod[j];
auto &offset = sub_lod.back();
size_t offset = sub_lod.back();
for (size_t k = 1; k < lod[j].size(); ++k) {
sub_lod.push_back(lod[j][k] + offset);
}
......
Markdown is supported
0% .
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