could you share your baseline?
I even could not run successfully on colab.
On kaggle, I ran up training. However, when evaluation, no module "ｍｍrotate.core" error.
Today, It became :
/opt/conda/lib/python3.7/site-packages/torch/utils/data/dataloader.py in _reset(self, loader, first_iter)
1146 # prime the prefetch loop
1147 for _ in range(self._prefetch_factor * self._num_workers):
-> 1148 self._try_put_index()
1149
1150 def _try_get_data(self, timeout=_utils.MP_STATUS_CHECK_INTERVAL):
/opt/conda/lib/python3.7/site-packages/torch/utils/data/dataloader.py in _try_put_index(self)
1380
1381 try:
-> 1382 index = self._next_index()
1383 except StopIteration:
1384 return
/opt/conda/lib/python3.7/site-packages/torch/utils/data/dataloader.py in _next_index(self)
669
670 def _next_index(self):
--> 671 return next(self._sampler_iter) # may raise StopIteration
672
673 def _next_data(self):
/opt/conda/lib/python3.7/site-packages/torch/utils/data/sampler.py in __iter__(self)
245 batch = [0] * self.batch_size
246 idx_in_batch = 0
--> 247 for idx in self.sampler:
248 batch[idx_in_batch] = idx
249 idx_in_batch += 1
/opt/conda/lib/python3.7/site-packages/mmdet/datasets/samplers/group_sampler.py in __iter__(self)
34 [indice, np.random.choice(indice, num_extra)])
35 indices.append(indice)
---> 36 indices = np.concatenate(indices)
37 indices = [
38 indices[i * self.samples_per_gpu:(i + 1) * self.samples_per_gpu]
<__array_function__ internals> in concatenate(*args, **kwargs)
ValueError: need at least one array to concatenate
Posted by: flydragon2023 @ Feb. 14, 2023, 1:58 a.m.Hi,
The baseline notebook was tested in Colab.
To execute the training in other environments, we recommend that you ensure the environment settings are similar to Colab (i.e, python version, pytorch version, mmrotate version etc)
If this solution doesn't work please let us know.
Shai,
MAFAT Challenge Team
Hello,
I encountered the same problem in Colab on the full dataset. when I tried to run it again it writes me:
RuntimeError: ReDet: ReResNet: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking argument for argument mat2 in method wrapper_bmm)
which is, as you written before, a problem that might be solved by rerunning the Colab.
what can we do?
Hi,
Please disconnect and delete runtime (at the runtime button in the toolbar) and re-run the notebook from the beginning.
We suspect it might have something to do with the installed packages dependecies, so you also must ensure you have installed all the required packages with the correct version.
Neta
MAFAT Challenge Team
Hi,
I tried to do both options but failed.It still shows me the same errors. when I run the notebook on colab with the minidataset it works but with the full dataset it doesn't.
Is it possible to upload new notebook where the full dataset works on colab?
Thanks!
Hi,
We are unable to reproduce the errors you mentioned.
We think it is caused by a failed download of the full dataset followed by an attempt to run the training cell multiple times.
A possible solution to the download problem is to download the data to your desktop and then upload it to your google drive. Then you can add to the notebook the following commands instead of the download commands:
from google.colab import drive
drive.mount('/content/drive')
!unzip '/content/drive/MyDrive/train_dataset.zip' -d '/content'
Shai,
MAFAT Challenge Team