problem loading data #824
Replies: 3 comments 1 reply
-
the memory requiremed is related to the input shape, but much more to the exact model you used (number a model parameter, and number of output) An other parameter is also the batch size as a batch size of 4 will take twice the memory size as a batch size of 2 ... (but ok , you already have a batch size of 1) |
Beta Was this translation helpful? Give feedback.
-
6G is a quite too low for 3DCNN ... |
Beta Was this translation helpful? Give feedback.
-
As @romainVala said, you probably won't be able to train with full volumes on a 6-GB GPU. I suggest you use patch-based training. |
Beta Was this translation helpful? Give feedback.
-
Hi,
Sorry as I am not sure I posted my question at the right place.
I am trying to load both training and validation set with len 3 from tio.suject with 4 images of 22.5mb.
I am trying to run that on 6Gb gpu but I can"t manage to do that and I don't know where I should look to improve my code.
Please find attached my code and the error.
I am thinking about to create a specific dataset but I am not really sure about that.
Moreoever I don't understand why the error states it tried to allocate 9gb whereas my files are much smaller and the dataset is very small.
Many thanks in advance
Beta Was this translation helpful? Give feedback.
All reactions