Hi
I am trying to solve this MNIST challenge using fastai.
I am running on Crestle.
Everything works fine up to the following:
data = ImageClassifierData.from_arrays(path=working_directory,
trn=(train_X, train_y),
val=(val_X, val_y),
bs = 32,
tfms=tfms_from_model(arch, sz),
classes=classes,
test=test_df
)
running that immidiately gives a big error, which is concluded by
PermissionError
I read on the forums that I should run some crestle related code, from lesson 1:
os.makedirs(‘data/dogscats/models’, exist_ok=True)
!ln -s /datasets/fast.ai/dogscats/train {PATH}
!ln -s /datasets/fast.ai/dogscats/test {PATH}
!ln -s /datasets/fast.ai/dogscats/valid {PATH}
os.makedirs('/cache/tmp', exist_ok=True)
!ln -fs /cache/tmp {PATH}
os.makedirs('/cache/tmp', exist_ok=True)
!ln -fs /cache/tmp {PATH}
This code is obviously specific, and won’t fit my case.
I don’t understand what this code does, why it solves the permission problem, or how to generalize to my case and folder structure.
Honestly, I also don’t like symbolic links. I prefer to know where the real everything is if possible.
How do I get past that permissions problem?
Thanks