首页解释RuntimeError: CUDA error: device-side assert triggered

解释RuntimeError: CUDA error: device-side assert triggered

时间: 2023-10-12 15:05:14 浏览: 224

pytorch模型提示超出内存RuntimeError: CUDA out of memory.

5星 · 资源好评率100%

跑模型时出现RuntimeError: CUDA out of memory.错误查阅了许多相关内容，原因是：GPU显存内存不够简单总结一下解决方法：将batch_size改小。取torch变量标量值时使用item()属性。可以在测试阶段添加如下代码： with torch.no_grad(): # 停止自动反向计算梯度参考： https://wisdomai.xyz/tool/pytorch/archives/2301 https://ptorch.com/news/160.html 作者：菜叶儿掉啦

RuntimeError: CUDA error: device-side assert triggered是一个CUDA错误，表示在GPU端触发了一些断言错误。这种错误通常发生在CUDA内核调用期间，并且可能由于异步报告错误而导致堆栈跟踪不准确。123 #### 引用[.reference_title] - *1* *2* [解决：RuntimeError: CUDA error: device-side assert triggered](https://blog.csdn.net/wotianxiu/article/details/130126332)[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^v92^chatsearchT0_1"}}] [.reference_item style="max-width: 50%"] - *3* [RuntimeError: CUDA error: device-side assert triggered](https://blog.csdn.net/yeeanna/article/details/129111182)[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^v92^chatsearchT0_1"}}] [.reference_item style="max-width: 50%"] [ .reference_list ]

阅读全文