打开微信,使用扫一扫进入页面后,点击右上角菜单,
点击“发送给朋友”或“分享到朋友圈”完成分享
【寒武纪硬件产品型号】必填*:MLU270
必填项,例如:MLU370
【使用操作系统】必填*:ubuntu
例如:ubuntu
【使用驱动版本】必填*: v4.8.0
【出错信息】必填*:
2022-11-03 17:05:28.671076: [cnrtError] [31174] [Card : 0] MLU unfinished. cnrtStream fail.
2022-11-03 17:05:28.671099: [cnrtError] [31174] [Card : 0] unknown errorCNRT ERROR in file "/pytorch/catch/torch_mlu/csrc/jit/fuser/fused_kernel.cpp" on line 271.
Traceback (most recent call last):
File "mlu_forward.py", line 227, in <module>
mlu_forward(opt.img, use_mlu=True, mlu220=save_offline_model)
File "mlu_forward.py", line 212, in mlu_forward
out = fuison_model(example_tensor)
File "/torch/venv3/pytorch/lib/python3.6/site-packages/torch/nn/modules/module.py", line 541, in __call__
result = self.forward(*input, **kwargs)
RuntimeError: CNRT error: Failure on Queue.
The above operation failed in interpreter, with the following stack trace:
【当前已做了哪些信息确认】选填:
1.CPU能运行,以及在PC端的pytoch1.7.0版可以进行torch.jit.trace().save(),可以进行模型量化并保存量化模型。
2.该网络中将后处理放入了网络中,之前出现错误"RuntimeError: Can not call cpu_data on an empty tensor",将后处理注释掉后(如果对后处理的Tensor加上(.to(ct.mlu_device()反而出现内存不停上涨卡死的情况),仅保留前向传播的变量部分便出现上述错误
【相关日志文档】选填
如有,可附件
【出错代码链接】选填:
该代码来自:https://github.com/TencentYoutuResearch/CrowdCounting-P2PNet
热门帖子
精华帖子