Crestle [easier alternative to AWS]

Worked perfectly! Thank you!

Thank you @anurag for doing this. I am trying it out now. where are you guys based at?

Just me. Based in SF.

1 Like

Iā€™m trying to extend the deeplearning1 notebook mnist.ipynb and am running into some bugs. On the line

lm.fit_generator(batches, batches.N, nb_epoch=1, validation_data=test_batches, nb_val_samples=test_batches.N)

I first had the error message

AttributeError: 'NumpyArrayIterator' object has no attribute 'N'

which is raised on this thread, with a solution but not an explanation. I changed .N to .n as suggested there and then had an exception on the same line:

error: identifier "cudnnSetFilterNdDescriptor_v4" is undefined

which seems to be a problem with incompatibility between different versions of the libraries, discussed on github.

Iā€™ve already been warned:

Using gpu device 0: Tesla K80 (CNMeM is enabled with initial size: 95.0% of memory, cuDNN 6021)
/home/nbuser/.local/lib/python2.7/site-packages/theano/sandbox/cuda/__init__.py:631: UserWarning: 
Your cuDNN version is more recent than the one Theano officially supports. 
If you see any problems, try updating Theano or downgrading cuDNN to version 5.1.
warnings.warn(warn)

Iā€™m wary of fooling around trying to change library versions when I donā€™t fully understand the crestle environment, and I donā€™t know the command to issue to downgrade cuDNN or to upgrade theano. @anurag, can you suggest a good way to work around this? Thanks!

Hi @anurag. Thanks for setting up this service, so far it seems great. Iā€™ve run into one problem however.

Iā€™m getting this error when running input 7 on lesson 1 notebook:



between the pictures posted, the notebook scrolls through the various errors in each line of the script, I but Iā€™ve just shown the first and last extremes of this.

Any idea what might be going on?
Thanks!

1 Like

Iā€™m on crestle and Iā€™m getting the same error. Itā€™s so far down in the libraries I canā€™t guess what the problem really is. Any help would be appreciated.

Thanks for the reports. I should have a fix out shortly.

Should be fixed if you restart Jupyter.

2 Likes

Thanks Anurag ā€“ appreciate the quick attention!

Hey @anurag first of all thank you very much for your contribution. Currently doing Lesson 1. I would like to know how or where are we reading the files on? Because we define path = ā€œ/datasets/fast.ai/dogscats/ā€ and I canā€™t seem to find that directory in the filesā€¦ So where should I store files if later on I would like to run code with my own dataset?

Thanks in advance.

Thank you so much!

Just out of curiosity, what was causing the problem?

Yes, thatā€™s fixed it for me, too, although I still have to change ā€œ.Nā€ to ā€œ.nā€.

Thanks!
Emrys
Whoops, I replied to @taguilera when I should have been replying to @anurag. Thanks, @anurag!

1 Like

@taguilera, /datasets/fast.ai/dogscats/ is at the whole filesystem root, as opposed to the per-user root that you see with the Jupyter file browser. If you hit new terminal in Jupyter, you get a linux shell. Type

ls /

to see the filesystem root and you should see datasets as one of the directories there, and you can cd around to explore. Type

ls ~

to see your own userā€™s files, which is what the Jupyter file browser displays.

If you store your own data anywhere in your own userā€™s files, it stays stored there. At least, mine does.

Good Luck!
Emrys

2 Likes

Hi Anurag,

I have just started using Crestleā€¦First of all I would like to thank you for making this. I find it very helpful. Just some doubt, is there any way to access file from local drivesā€¦

Thnxā€¦

I would transfer data to a public host and use wget to download it in the terminal.

I upgraded to cuDNN 6, which Theano doesnā€™t like. Downgraded back to v5. This should all be fixed in the next version of Part 1 which does not use Theano (which has ceased active development anyway).

Ahhh I see.

thnx a lotā€¦

anyone know how i can get kaggle data into this? iā€™m trying the Statefarm distracted driver competition and the data is 4gb. https://www.kaggle.com/c/state-farm-distracted-driver-detection/data

Edit: Just noticed that the Statefarm data is already there in the kaggle datasets folder - awesome! Only problem is it seems to be in zipped folders and I canā€™t seem to find the train and validation sets.

Thanks again for everything youā€™ve done. Truly couldnā€™t be more grateful.

There was an issue with kaggleā€™s statefarm data earlier. Itā€™s fixed and everything is unzipped under /datasets/kaggle/state-farm-distracted-driver-detection. Hope this helps!

1 Like