未验证 提交 5ffb48d6 编写于 作者: X Xin Pan 提交者: GitHub

Merge pull request #14073 from panyx0718/fix10

avoid blocking everyone
......@@ -156,12 +156,6 @@ ParallelExecutor::ParallelExecutor(
params, member_->local_scopes_, member_->use_cuda_);
#endif
// If the loss_var_name is given, the number of graph should be only one.
if (loss_var_name.size()) {
PADDLE_ENFORCE_EQ(ir::GraphNum(*graph), 1,
"The number of graph should be only one");
}
if (exec_strategy.type_ == ExecutionStrategy::kDefault) {
member_->executor_.reset(new details::ThreadedSSAGraphExecutor(
exec_strategy, member_->local_scopes_, places, std::move(graph)));
......
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