Reputation: 51
I am training my model with colab. I have encountered this error more than once, when I use the enumeration function in colab to iterate my training data on Google Drive. The time when the error occurred is uncertain. The error may be after training a few epoch, or it may happen in several iterations.
I tried to modify the path of the dataset in Google Drive, or reduce the batchsize, these have no effect.
192 batch_time = AverageMeter()
193 losses = AverageMeter()
--> 194 for i, sample in enumerate(train_loader):
195 start = time.time()
196 input = sample['image'].cuda()
File "/usr/local/lib/python3.6/dist-packages/PIL/Image.py", line 2530, in open
fp = builtins.open(filename, "rb")
OSError: [Errno 5] Input/output error: './nyud_cropped/train_labels/001187.png'
Upvotes: 5
Views: 2166
Reputation:
the same error appeared to me when I executed the copy command that it can't access the folder that contains the data set due to I_O error and google drive had time out
Upvotes: 0
Reputation: 38579
My recommendation is to first copy data out of Drive to the ephemeral VM machine before training. That may also speed things up.
Upvotes: 3