CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace belo

CUDA kernel errors可能会在其他API调用时以异步方式报告，因此下面的堆栈跟踪可能不正确。为了调试，可以考虑传递CUDA_LAUNCH_BLOCKING=1。这意味着在GPU上运行程序时可能会出现报错：RuntimeError：CUDA error。由于代码在主机和设备之间是异步执行的，当设备抛出错误时，主机可能正在执行一些其他操作，这可能导致堆栈跟踪不准确。<span class="em">1</span><span class="em">2</span><span class="em">3</span>
#### 引用[.reference_title]
- *1* *2* *3* [CUDA_LAUNCH_BLOCKING=1的作用](https://blog.csdn.net/weixin_43590796/article/details/131588761)[target="_blank" data-report-click={"spm":"1018.2226.3001.9630","extra":{"utm_source":"vip_chatgpt_common_search_pc_result","utm_medium":"distribute.pc_search_result.none-task-cask-2~all~insert_cask~default-1-null.142^v93^chatsearchT3_2"}}] [.reference_item style="max-width: 100%"]
[ .reference_list ]