Issues when running the code in Colab

[Jose-Augusto-C-M]

Hello, first I would like to thank you guys for all the hard work put in mmlab in general, It´s an amazing contribution.

For the error, I´m try to run the codes: '[MMDet_InstanceSeg_Tutorial.ipynb] and [MMDet_Tutorial.ipynb]' on colab, in both I´m facing an error when running this part of the code:

'
from mmdet.datasets import build_dataset
from mmdet.models import build_detector
from mmdet.apis import train_detector

Build dataset

datasets = [build_dataset(cfg.data.train)]

Build the detector

model = build_detector(cfg.model)

Add an attribute for visualization convenience

model.CLASSES = datasets[0].CLASSES

Create work_dir

mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
train_detector(model, datasets, cfg, distributed=False, validate=True)
'

The result of the code is:
'

loading annotations into memory...
Done (t=0.01s)
creating index...
index created!

AttributeError Traceback (most recent call last)
in ()
15 # Create work_dir
16 mmcv.mkdir_or_exist(osp.abspath(cfg.work_dir))
---> 17 train_detector(model, datasets, cfg, distributed=False, validate=True)

2 frames
/content/mmdetection/mmdet/apis/train.py in train_detector(model, dataset, cfg, distributed, validate, timestamp, meta)
161 find_unused_parameters=find_unused_parameters)
162 else:
--> 163 model = build_dp(model, cfg.device, device_ids=cfg.gpu_ids)
164
165 # build optimizer

/usr/local/lib/python3.7/dist-packages/mmcv/utils/config.py in getattr(self, name)
511
512 def getattr(self, name):
--> 513 return getattr(self._cfg_dict, name)
514
515 def getitem(self, name):

/usr/local/lib/python3.7/dist-packages/mmcv/utils/config.py in getattr(self, name)
47 else:
48 return value
---> 49 raise ex
50
51

AttributeError: 'ConfigDict' object has no attribute 'device'
'

This happens when using the exact code from the demo, I did not change anything. What should I do?

[12567878]

I came up with the same trouble.
In mmdet/apis/train.py:163 change it to:
model = build_dp(model, device_ids=cfg.gpu_ids) # delete parameter cfg.device,it will work.

API:def build_dp(model, device='cuda', dim=0, *args, **kwargs):
Passing cfg.device='cuda' may work,too.

[Jose-Augusto-C-M]

Thank you bro or sister 😅, I will come back to investigate the issues with my code, there's some time that i don't revisit this demand. But very fucking thank you for you help. 🖖

…

On Wed, Apr 27, 2022, 19:56 12567878 ***@***.***> wrote: I came up with the same trouble. In mmdet/apis/train.py:163 change it to: model = build_dp(model, device_ids=cfg.gpu_ids) # delete parameter cfg.device,it will work. API:def build_dp(model, device='cuda', dim=0, *args, **kwargs): Passing cfg.device='cuda' may work,too. — Reply to this email directly, view it on GitHub <#7841 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AIZQTU6U4LOYE6FVDQPBYN3VHF5WBANCNFSM5UODPWRQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[ai21mtech13002]

Hi, I cam across the same issue. I was bale to resolve it be setting
cfg.device='cuda' in my configuration script