How to use best.pt that has been obtained from training yolov5s model?

[mr-mainak]

Hi,
Currently when I load the weights I get the following error:

AttributeError: Can't get attribute 'DetectionModel' on <module 'models.yolo' from '/home/_/_/python/onnx_env/lib/python3.8/site-packages/yolort/v5/models/yolo.py'>

I'm attaching the total code bolow:

model_path = "/local/path/to/the/weights/best_v2.0.pt"
onnx_path = "/local/path/to/the/weights//best_v2.0.onnx"

model = YOLOv5.load_from_yolov5(
    model_path,
    size=size,
    size_divisible=size_divisible,
    score_thresh=score_thresh,
    nms_thresh=nms_thresh,
)

model = model.eval()
model = model.to(device)

This is a in a different python environment from the one that i've trained my model, i.e. I've trained my model in env1 and performing the inference in env2.
Any suggestions on where I'm wrong is highly appreciated.
Regards

[mr-mainak]

@zhiqwang The above problem is solved by changing the path_ultralytics_yolov5 in helper.py. However after loading the model when I'm performing conversion export_onnx(model=model, onnx_path=onnx_path, opset_version=opset_version) this error is coming:

AttributeError                            Traceback (most recent call last)
[/home/mainak/ms/python/yolov5_inference/opencv_inf/onnx/convert.ipynb](https://file+.vscode-resource.vscode-cdn.net/home/mainak/ms/python/yolov5_inference/opencv_inf/onnx/convert.ipynb) Cell 7 line 1
----> [1](vscode-notebook-cell:/home/mainak/ms/python/yolov5_inference/opencv_inf/onnx/convert.ipynb#W6sZmlsZQ%3D%3D?line=0) export_onnx(model=model, onnx_path=onnx_path, opset_version=opset_version)

File [~/ms/python/onnx_env/lib/python3.8/site-packages/yolort/runtime/ort_helper.py:60](https://file+.vscode-resource.vscode-cdn.net/home/mainak/ms/python/yolov5_inference/opencv_inf/onnx/~/ms/python/onnx_env/lib/python3.8/site-packages/yolort/runtime/ort_helper.py:60), in export_onnx(onnx_path, checkpoint_path, model, size, size_divisible, score_thresh, nms_thresh, version, skip_preprocess, opset_version, batch_size)
     24 """
     25 Export to ONNX models that can be used for ONNX Runtime inferencing.
     26 
   (...)
     44         when inferring if you want to export multiple batches ONNX models. Default: 1
     45 """
     47 onnx_builder = ONNXBuilder(
     48     checkpoint_path=checkpoint_path,
     49     model=model,
   (...)
     57     batch_size=batch_size,
     58 )
---> 60 onnx_builder.to_onnx(onnx_path)

File [~/ms/python/onnx_env/lib/python3.8/site-packages/torch/utils/_contextlib.py:115](https://file+.vscode-resource.vscode-cdn.net/home/mainak/ms/python/yolov5_inference/opencv_inf/onnx/~/ms/python/onnx_env/lib/python3.8/site-packages/torch/utils/_contextlib.py:115), in context_decorator.<locals>.decorate_context(*args, **kwargs)
    112 @functools.wraps(func)
    113 def decorate_context(*args, **kwargs):
    114     with ctx_factory():
--> 115         return func(*args, **kwargs)
...
--> 146         val = img.shape[-2:]
    147         assert len(val) == 2
    148         original_image_sizes.append((val[0], val[1]))

AttributeError: 'NoneType' object has no attribute 'shape'

Any suggestions? Also I'm in v7.0 of yolov5 repo.
Regards

[mr-mainak]

It's already raised as an issue #485...just followed the steps. Really great repo.