FYI folks, watching our demand go up during the current lesson we’ve added extra capacity! :]
is it possible to use code-server to run VS Code on a paperspace gradient remote server?
Another question, sorry.
Is there a possibility to run JupyterLab on paperspace gradient? JupyterLab released new visual debugger which is very cool. Would be great to have that option on paperspace platform! thanks!
@miwojc Absolutely! Many of our pre-configured templates run Jupyter Lab eg the PyTorch and TensorFlow 2.0 templates. If you’d like to start with your own, just head to the custom container tab and run
jupyter lab --allow-root --ip=0.0.0.0 as the command (make sure Jupyter Lab is installed of course ). Here’s an example:
Hope that helps!
@miwojc I just tested VS Code on our Core platform (virtual machines) and it worked! It’s pretty easy to set up – just be sure to add a public IP address. Feel free to DM me with any questions.
For VS Code to work on Gradient, we would need to create an option to run launch this service (where we normally run the Jupyter service), securely expose the port, and add a button to open VS Code in the browser in our UI. It’s not a huge project and I like the idea though we would need to do a bit of research to see what the demand is like.
Thank you for the instructions. I have managed to start a new notebook with JupyterLab instead of Jupyter Notebooks. I installed visual debug extension and xeus-python
jupyter labextension install @jupyterlab/debugger
conda install xeus-python -c conda-forge. I also instaled conda and nodejs and npm as per instructions. However i didn’t have option to selectd xeus kernel notebook. At this moment that’s the only option to use visual debugger. Also i restarted (stopped and started) the notebook but i kind of lost conda by doing this, so couldn’t activate my environment. I must be making a lot of mistakes there. Any help will be much appreciated! thanks!
@miwojc One of our engineers hacked this together with very minimal effort:
We definitely need productize this
this looks GREAT!!!
edit: here’s how binder did it, if that helps…
edit_1: and great thing also is that jupyter notebooks are nicely supported by VS Code as well.
I tried something like that with the predefined pytorch container which runs lab but failed at installing xeus.
Having a container like the “Paperspace + Fast.AI 2.0 (V4)” but with jupyter lab and the debugger installed would be super helpful for everyone @dkobran can you help to make that happen?
It has been a while that I haven’t log into Paperspace. In the billing tab, I found that there are 3 different kinds of credit: Gradient Credit, Referral Credit and credit (in the top right). And the credit in the top right is not the sum of these 2 others ones. Can someone help me to clarify what are these credits ?
Im having the problem that some files are being deleted. For instance I downloaded three folder of images yesterday and now one of them is gone…Also utils.py is getting deleted daily. Happen to know whats going on?
Im shutting down the notebooks and the instance.
I havent been hitting Logout or Quit in Jupyter. Maybe thats the reason? Or should I be git committing changes?
Did you save your data in /storage ? As I read from PaperSpace that just the data in /storage will be saved permanently
I saved my images in the same directory as the notebooks. I tried saving it in storage but didnt know how to retrieve the files when training my model. I tried /storage but it looks for a folder called storage in the current folder
You should be able to use a relative path to get to the images folder in storage. I used the following for the 02_production notebook, saving my images in the storage area, and was able to run the rest of the notebook normally.
path = Path(’…/…/storage/data/your_image_dataset’)
Hope it helps.
Edit: I also had issues when starting up my notebook in a different type of instance (it doesn’t default to P5000-free if there are none available), now that I’m ensuring that I always have the same instance type it all seems to be ok.
What does it mean when trival operations (like
print(1)) take 30 seconds or more to execute in a Paperspace notebook? I’m using a paid instance, so I don’t think I’m being preempted.
I think it might depends on your network bandwidth. If you have a slow internet so it take time to send the data from your browser to Paperspace and get back the result
What is the good way to use terminal in Paperspace ? All the shortcut for terminal in linux doesn’t work.
And I encounter a lag when using terminal in Paperspace, I think it is because of the slow internet I have so I think I can’t do much things related to this. But I ask anyway if someone have good way to deal with, maybe run all the linux command in the notebook so we don’t see the lag. Thanks
which shortcuts are you talking about?
i switch terminal to bash by typing
bash in the terminal, then i can use up arrow to get last command for example
Not sure, but I am having hard time also changing the theme of the jupyter notebook on paperspace
do you want dark theme for JN? i use dark reader browser extension, works quite well.