提交 e33b5e43 编写于 作者: M mindspore-ci-bot 提交者: Gitee

!3633 fix dataset & train gil lock of gpu process

Merge pull request !3633 from panfengfeng/fix_dataset_train_gil_of_gpu
......@@ -281,7 +281,10 @@ py::tuple GPUSession::RunOp(const OpRunInfo &op_run_info, const GraphInfo &graph
RunOpAllocateMemory(input_tensors, kernel_graph.get());
// Execute the computation
LoadInputData(kernel_graph, input_tensors);
Execute(kernel_graph);
{
py::gil_scoped_release gil_release;
Execute(kernel_graph);
}
// Fetch outputs
VectorRef outputs;
UpdateOutputs(kernel_graph, &outputs, input_tensors);
......
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