yolov8 + nebuly | AttributeError: type object 'DummyClass' has no attribute 'models'

[scraus]

I tried reproducing the results in this notebook but I got the following errors:

[saim212]

I'm facing the same error any solution?

[vtl-btz]

Got the same problem.
As a workaround, you can use the previous speedster version
!pip install speedster==0.2.1

also, I can't find the latest speedster code in this repository. Maybe it exists only in PyPi

I'm facing the same error any solution?

[saim212]

Thanks for the response. now facing this issue.

     97         pred = [pred.cpu()]
     98     else:
---> 99         pred = [p.cpu() for p in pred]
    100     return pred
    101 

AttributeError: 'list' object has no attribute 'cpu'

Got the same problem. As a workaround, you can use the previous speedster version !pip install speedster==0.2.1

also, I can't find the latest speedster code in this repository. Maybe it exists only in PyPi

I'm facing the same error any solution?

[valeriosofi]

Hello, thank you for using speedster! We have just released speedster 0.4.0, where this issue is solved! Please update your local version by running pip install -U speedster and let me know if you have any other issues 😁

[saim212]

with speedster==0.4.0 still facing the same issue.

[saim212]

python -m nebullvm.installers.auto_installer --compilers all
the above command failed to install the following dependencies.

2023-06-19 05:30:44 | WARNING  | Unable to install torch_tensor_rt on this platform. The compiler will be skipped. 
2023-06-19 05:30:44 | INFO     | Trying to install faster_transformer on the platform...
2023-06-19 05:30:44 | WARNING  | Unable to install faster_transformer on this platform. The compiler will be skipped. 

[valeriosofi]

Hi @saim212, Can you share the notebook that you are using? I've just tried with the official one and It works correctly, you can try it here: YOLOv8 Notebook

[valeriosofi]

Hi @vtl-btz, you can find the speedster code here :)

Got the same problem. As a workaround, you can use the previous speedster version !pip install speedster==0.2.1

also, I can't find the latest speedster code in this repository. Maybe it exists only in PyPi

I'm facing the same error any solution?

[saim212]

Hi @saim212, Can you share the notebook that you are using? I've just tried with the official one and It works correctly, you can try it here: YOLOv8 Notebook

Hello @valeriosofi, I tested your notebook. It's working fine but I was trying to optimize the yolov8s-pose model. There I faced this issue.

[saim212]

For Yolov8-Pose model:

class YOLOWrapper(torch.nn.Module):
    def __init__(self, yolo_model):
        super().__init__()
        self.model = yolo_model.model
    
    def forward(self, x, *args, **kwargs):
        res = self.model(x)
        return res[0], *tuple(res[1][0])
        
model_wrapper = YOLOWrapper(yolo)

class OptimizedYOLO(torch.nn.Module):
    def __init__(self, optimized_model):
        super().__init__()
        self.model = optimized_model
    
    def forward(self, x, *args, **kwargs):
        res = self.model(x)
        return res[0], list(res[1][0:])
    
optimized_wrapper = OptimizedYOLO(optimized_model)