关于 yolov5s.torchscript.pt

yulin1029 commented 3 months ago

您好，我想请教一下您是如何获得的yolov5s.torchscript.pt，我利用export.py运行命令将yolov5s.pt转换为yolov5s.torchscript.pt，转换后不能正常使用，报错内容放在最后了。如果方便的话能提供获得yolov5s.torchscript.pt的环境或者需要修改的参数及注意事项吗，不知道您使用的是否是yolo官方自带的模型转换的代码export.py。报错如下：terminate called after throwing an instance of 'std::runtime_error' what(): The following operation failed in the TorchScript interpreter. Traceback of TorchScript, serialized code (most recent call last): File "code/torch/models/yolo.py", line 48, in forward _37 = (_4).forward(_25, _36, ) _38 = (_2).forward((_3).forward(_25, _37, ), _31, ) _39 = (_0).forward(_35, _37, (_1).forward(_38, ), )


    return (_39,)
class Detect(Module):
  File "code/__torch__/models/yolo.py", line 74, in forward
    _48 = torch.split_with_sizes(torch.sigmoid(_47), [2, 2, 2], 4)
    xy, wh, conf, = _48
    _49 = torch.add(torch.mul(xy, CONSTANTS.c0), CONSTANTS.c1)
          ~~~~~~~~~ <--- HERE
    xy0 = torch.mul(_49, torch.select(CONSTANTS.c2, 0, 0))
    _50 = torch.pow(torch.mul(wh, CONSTANTS.c0), 2)

Traceback of TorchScript, original code (most recent call last):
/home/cqut/yolov5/models/yolo.py(76): forward
/home/cqut/.local/lib/python3.6/site-packages/torch/nn/modules/module.py(1039): _slow_forward
/home/cqut/.local/lib/python3.6/site-packages/torch/nn/modules/module.py(1051): _call_impl
/home/cqut/yolov5/models/yolo.py(123): _forward_once
/home/cqut/yolov5/models/yolo.py(212): forward
/home/cqut/.local/lib/python3.6/site-packages/torch/nn/modules/module.py(1039): _slow_forward
/home/cqut/.local/lib/python3.6/site-packages/torch/nn/modules/module.py(1051): _call_impl
/home/cqut/.local/lib/python3.6/site-packages/torch/jit/_trace.py(959): trace_module
/home/cqut/.local/lib/python3.6/site-packages/torch/jit/_trace.py(744): trace
export.py(122): export_torchscript
export.py(106): outer_func
export.py(579): run
/home/cqut/.local/lib/python3.6/site-packages/torch/autograd/grad_mode.py(28): decorate_context
export.py(667): main
export.py(672): <module>
RuntimeError: The size of tensor a (48) must match the size of tensor b (80) at non-singleton dimension 2

yulin1029 commented 3 months ago

我也尝试在csdn上下载yolov5.torchscript.pt，依旧会报错，期待收到您的信息

ColdZera97 commented 3 months ago

你好，我现在也在跑这个程序，请问你是怎么解决段错误的这个问题？？？我搞不定，卡很久了，方便的话，可以加个微信吗？一起交流学习，我是做slam的,没有师兄，老师也不管，感觉很艰难。。微信18238489393

YWL0720 / YOLO_ORB_SLAM3

关于 yolov5s.torchscript.pt #31