RuntimeError: DataLoader worker (pid 5657)

最新推荐文章于 2026-04-28 09:03:03 发布

原创最新推荐文章于 2026-04-28 09:03:03 发布 · 1.6w 阅读

7 ·

本内容遵循CC 4.0 BY-SA版权协议

在使用PyTorch的DataLoader进行大规模数据集读取时，遇到内存溢出导致进程被信号杀死的问题。通过调整num_workers参数和batch-size，以及检查系统内存使用情况，可以有效解决此问题。

Qwen3-32B-Chat 私有部署镜像 | RTX4090D 24G 显存 CUDA12.4 优化版

Qwen

文本生成

Qwen3

本镜像基于 RTX 4090D 24GB 显存 + CUDA 12.4 + 驱动 550.90.07 深度优化，内置完整运行环境与 Qwen3-32B 模型依赖，开箱即用。

RuntimeError: DataLoader worker (pid 5657) is killed by signal: Killed. Details are lost due to multiprocessing. Rerunning with num_workers=0 may give better error trace.

I encountered into the following problem when I was trying to read a batch （32）of relatively large data sample with multi-threaded DataLoader (with num_workers=8 for example).
I have tried increasing the shared memory of ubuntu but did not work.
It will run without the num_workers argument, but it is too slow to learn from a large data set with single thread.

参考了一些答案后改成num-worker为4 batch-size为20

In my case, I find that it caused by out of memory of machine(note that , not GPU memory). To test that when run your code, you can use command top to see the whether the memory and swap memory are used up. If so , it is the memory problem. You can use a small batch, use a small num_workers, expand your swap memory,etc. Any way could decrease your memory load would be helpful.