Platform: Paperspace (Free; Paid options)

when trying to load the storage directory i get:

[Errno 2] No such file or directory: ‘…/…/storage/data’

code:
path = Path(’…/…/storage/data’)
path.ls(file_exts=‘export.pkl’)

Happen to know what Im doing wrong?

Not sure why I had three dots in there, my bad. Does

do you mean it was supposed to be 2 dots?

and no this doesnt work
path = Path(’…/…/storage/data’)
path.ls(file_exts=‘export.pkl’)

ive also tried
path = Path(…/…/storage/data)
path = Path(…/…/‘storage/data’)
path = Path(…/…/‘storage’/‘data’)

Yes two dots, for some reason typing that in markdown changes it to three dots.

The best thing for you to do is open a terminal (Jeremy showed how to do it in one of the lectures) and navigate to your data folder - use:

cd directory_name/

to move into a directory. And (two dots):

cd …/

To move up a directory. Figure out how you move from the directory with your notebook to the data directory, and then recreate that in your path string. ‘Surround the whole string’ like you were doing first.

‘…/…/storage/data’ means “go up two directories, then into the storage directory, and then into the data directory”. I guess that your problems is that this is not the actual location of your data directory for some reason.

1 Like

yay! thanks so much for your patience! I got it working with:

path = Path(’…/…/data’)

2 Likes

@paulsolomon I had the same experience. I ended up writing some of the documentation for Returning to Work with Gradient . Note the large orange arrows and the bold text at the top. :slight_smile:

What is intuitive to us from the physical world doesn’t apply: we figure that that notebook lives on a machine, so if you’ve selected a machine, that’s where your notebook sits. But that isn’t the case. The notebook is what persists (though there are currently weird problems with that in paperspace) and you can slide any machine, so to speak, in under that notebook.

1 Like

Just to expand on @dhoa’s reply: you shouldn’t need to. Remember that you should be making a copy of whatever notebook you’re working on. The reason for this is that if you do a git pull to see if there is an update from the repo, you’ll get a warning that any file that you’ve modified from the repo is about to get overwritten. If you’ve modified that file with content or experiments of your own, it will indeed get overwritten unless you do things like git stash, which is a way of temporarily ‘hiding’ something from git while it does its thing; you can then unhide it (‘pop’ is actually the command) after you’ve done your pull. Far simpler to make a copy, with a different name like ‘notebook X copy’ which won’t get overwritten.

You’re probably not really interested in keeping track of your copy. All you’re interested in is that it remains there for you to use another day. That should happen if you’ve got it in a place where paperspace lets it persist.

3 Likes

thank you for the explanation. I was curious about this because random folders and files were being deleted and it was getting super annoying having to reupload them everytime i start up an instance.

I ended up moving the folder to /storage and now its looks, knock on all the wood in the area, it will stay

1 Like

did you get an answer to this one?

Is anyone facing issue with Paperspace where platform is failing to even recognize core functions? Any help is greatly appreciated. Please note I have tried to close and reopen the session.

Hey everyone!
I’ve only recently started using Paperspace Gradient Notebooks for my personal DL work. And while the experience is great, I was wondering if anyone has tried using Tensorboard with the service?
According to this: https://docs.paperspace.com/gradient/tensorboards/using-tensorboards, this is a feature just for Gradient Enterprise customers which I find a bit problematic because Gradient Enterprise is something most of us won’t ever be signing up for using in our personal work. And tensorboard is kind of supposed to be free and open-source :sweat_smile:

I tried creating a tensorboard inside my Jupyter notebook by using the %tensorboard magic command but that gave a server error. Also, upon trying to start it through a terminal and accessing it over the 6006 port (using the Instance’s public API which I got through the terminal) , the same nginx server error came up.

Has anyone had success here?
@dkobran @tomg can you help out here?

2 Likes

@akashpalrecha Great question. TL;DR Everything in the enterprise version (a customer-managed version that can run on-premise) will make it into our SaaS/hosted version soon.

When we started building this customer-managed version, we had to rebuild a few parts to support on-premise environments. Some of this work is currently incompatible with our SaaS/hosted version which was built with certain assumptions around multi-tenancy, managing the compute/storage clusters, etc. The fact that TensorBoards are only included in the enterprise version is merely the unfortunate but necessary result of having two separate code streams that will ultimately merge.

Sorry about this. We are open sourcing our installer and making it very easy to deploy the customer-managed version. We intend to offer a free plan with this version. I’m not 100% sure if this will be well-suited for personal projects but you’re more than welcome to try as soon as it lands :slight_smile:

2 Likes

I am unable to save my epoch values on Paperspace. Has anyone else seen this?

@vikbehal
If you read the last line in your error, it says that the file system is read only. And that’s expected as it’s not the location where paperspace gives you your own storage.
You simply need to change the save location to your writable storage directory.

Yup. Any thoughts on how can I pass a path to my learn.export() object?

I understand. Thanks for the detailed reply. Are there any possible workarounds to this ?

you can simply set learn.path to any directory you want where it will then save your model.
Simply: learn.path = XYZ

Thank you. For some reason it still does not work :confused:

learn.path = Path('notebooks/course-v4/nbs')
This should fix it.