yolov5离线推理报错

# 一、问题现象（附报错日志上下文）：
使用离线推理脚本提示size not match，从onnx导出om模型时，有设置batch-sz=1, 为何在推理时输入的大小是2457600 （2\*3\*640\*640）？
```
[INFO] Created a temporary directory at /tmp/tmph9df77th
[INFO] Writing /tmp/tmph9df77th/_remote_module_non_sriptable.py
/usr/local/python3.7.5/lib/python3.7/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension: libc10_cuda.so: cannot open shared object file: No such file or directory
  warn(f"Failed to load image Python extension: {e}")
[INFO] acl init success
[INFO] open device 0 success
[INFO] load model nanji_bs1_aipp.om success
[INFO] create model description success
./datasets/nanji-test
Scanning 'datasets/nanji-test/valid/labels_yolov5_v6.cache' images and labels... 288 found, 0 missing, 1 empty, 0 corrupted: 100%|████████████████████████████████████| 288/288 [00:00<?, ?it/s]
  0%|                                                                                                                                                                   | 0/288 [00:00<?, ?it/s]
[ERROR] Check i:0 name:images in size:2457600 needsize:4915200 not match
[ERROR] Check InVector failed ret:-1
  0%|                                                                                                                                                                   | 0/288 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "om_val.py", line 78, in <module>
    main(opt, cfg)
  File "om_val.py", line 44, in main
    pred_results = forward_nms_script(model, dataloader, cfg)
  File "/home/tangwg/code/modelzoo-GPL/built-in/PyTorch/Official/cv/object_detection/Yolov5_for_PyTorch_v6.0/common/util/model.py", line 66, in forward_nms_script
    result = model.infer([img])
  File "/usr/local/python3.7.5/lib/python3.7/site-packages/ais_bench/infer/interface.py", line 510, in infer
    return self.run(inputs, out_array)
  File "/usr/local/python3.7.5/lib/python3.7/site-packages/ais_bench/infer/interface.py", line 450, in run
    outputs = self.session.run(self.outputs_names, inputs)
RuntimeError: [-1][ACL: general failure] 
[INFO] unload model success, model Id is 1
[INFO] end to reset device 0
[INFO] end to finalize acl
```

# 二、软件版本:
-- CANN 版本 (e.g., CANN 8.0.RC1.alpha003 Ascend910B)
--Pytorch 版本: 1.11.0
--Python 版本 (Python 3.7.5):
--操作系统版本 (e.g., CentOS 7.6):
![输入图片说明](https://foruda.gitee.com/images/1717642750386162703/074957fa_8916109.png "屏幕截图")

# 三、测试步骤：
## 1. 导出ONNX
python3 export.py --data ./data/kitchen-web.yaml --weights ./runs/train/nanji/weights/last.pt
```
export: data=./data/kitchen-web.yaml, weights=./runs/train/nanji/weights/last.pt, imgsz=[640, 640], batch_size=1, device=cpu, half=False, inplace=False, train=False, optimize=False, int8=False, dynamic=False, simplify=False, opset=13, topk_per_class=100, topk_all=100, iou_thres=0.45, conf_thres=0.25, include=['onnx']
YOLOv5 🚀 2024-5-28 torch 1.11.0+cpu CPU

Fusing layers... 
Model Summary: 290 layers, 20856975 parameters, 0 gradients
/usr/local/python3.7.5/lib/python3.7/site-packages/torch/functional.py:568: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at  ../aten/src/ATen/native/TensorShape.cpp:2228.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]

PyTorch: starting from runs/train/nanji/weights/last.pt (83.9 MB)

ONNX: starting export with onnx 1.14.1...
/home/tangwg/code/modelzoo-GPL/built-in/PyTorch/Official/cv/object_detection/Yolov5_for_PyTorch_v6.0/models/yolo.py:61: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  if self.grid[i].shape[2:4] != x[i].shape[2:4] or self.onnx_dynamic:
WARNING: The shape inference of prim::Constant type is missing, so it may result in wrong shape inference for the exported graph. Please consider adding it in symbolic function.
WARNING: The shape inference of prim::Constant type is missing, so it may result in wrong shape inference for the exported graph. Please consider adding it in symbolic function.
WARNING: The shape inference of prim::Constant type is missing, so it may result in wrong shape inference for the exported graph. Please consider adding it in symbolic function.
WARNING: The shape inference of prim::Constant type is missing, so it may result in wrong shape inference for the exported graph. Please consider adding it in symbolic function.
WARNING: The shape inference of prim::Constant type is missing, so it may result in wrong shape inference for the exported graph. Please consider adding it in symbolic function.
WARNING: The shape inference of prim::Constant type is missing, so it may result in wrong shape inference for the exported graph. Please consider adding it in symbolic function.
ONNX: export success, saved as runs/train/nanji/weights/last.onnx (83.9 MB)
ONNX: run --dynamic ONNX model inference with: 'python detect.py --weights runs/train/nanji/weights/last.onnx'

Export complete (4.54s)
```
## 2. 转换om
 atc --model=nanji.onnx --output=nanji_bs1_aipp --framework=5 --input_format=NCHW --soc_version=Ascend910B --input_shape="images:1,3,640,640"

## 3. 推理
python3 om_val.py --tag 6.0 --model=nanji_bs1_aipp.om --nms_mode nms_script --data_path ./datasets/nanji-test

Ascend/samples
暂停

内容风险标识

评论 (1)

Ascend/samples暂停 .gitee-modal { width: 500px !important; }

内容风险标识