It’s stopped! I’m not sure if it’s a very-speedy admin, or if it just fixed itself within minutes of my posting, but either way: I’m good. Thanks.
Hi Jeremy, could you please say how to push ssh credentials using a terminal in Jupyter as you suggested before? I am having the same error “Network error: Unexpected token < in JSON at position 0”.
Second question: If I cannot setup an SSH connection with Salamander, how can I download my files from Salamander?
Third question: Is it thread an official support page of Salamander? and is here the appropriate place to ask about Salamander bugs and fixes?
You can paste ssh credentials directly into a jupyter terminal.
The easiest way to download files is often to use the ipython FileLink function.
There isn’t any support provided with Salamander, although I’m happy to answer questions here on the forum when I can.
1 Like
Does anyone know how to use the Redeem AWS Coupon?
the link on this page https://salamander.ai/redeem-aws-coupon tell me to have a github education. When I have it, I apply for the aws education.
And now, I do not know what to do.
Can anybody help me?
Morning, all! I’m having an issue with Salamander - every time I launch my server, I get the following message while waiting for it to start up:
Waiting for Jupyter Notebook to start…
Traceback (most recent call last):
File “/home/ubuntu/anaconda3/bin/jupyter”, line 7, in
from jupyter_core.command import main
ModuleNotFoundError: No module named ‘jupyter_core’
Any ideas what I can do to resolve?
I say this with the greatest hesitation and the utmost respect for fastai and it’s mission, but Salamander is just not working, at least for me. I am now getting 500 Internal Server Error on every notebook, which was not happening a week ago. I have updated the course and done a gitpull. Every time I create a new server instance, it works for a while and then stops working. I have spent many hours now trying to get it to work, ssh in, etc.
I don’t know how much of this is just me (although I am not doing anything unique) and how much this affects other users, but I’m leaving this post in case there are other users with the same problem so that you know you are not alone.
Google Collab is a free option that has worked more consistently for me than Salamander. I also have an Azure instance set up that may be working but haven’t really used it yet. I have had a lot of trouble getting GCloud to work with the instructions given. I am spending so much time trying to get servers to work that I have not made much actual progress with the AI part.
Again, this is not to diminish the extraordinary effort of the fastai team. I have learned a lot and think the course is great. I just wish I wasn’t spending so much time trying to get a server to work.
@minh If you open a Jupyter Notebook and go to “File” you will see an option to download your notebook. This is by far the easiest way to do it and does not require ssh, which I never managed to get to work.
Added: Google Cloud is now working for me. A week ago I could not create the instance despite having an adequate GPU quota. So you can try that if Salamander is not working.
2 Likes
Thanks Mark. I have completely switched to AWS which is far more reliable and I find it easy to use too. As a student I received $150 AWS credit. It is more than enough to finish the deep learning course. I can’t believe a service like Salamander who is taking user money while providing no support whatsoever. Ridiculous!
2 Likes
I’m also having trouble accessing notebooks in Salamander, getting error 500 after a while. I tried destroying the old and creating a new server, and it worked, but after a while the same is happening. If I go to /lab instead of /tree I can open the notebooks but the kernel starts up.
Can someone help??
1 Like
Sorry to reply to my own post, but anyone have any ideas? If I destroy the server and recreate, I assume I lose my data, correct?
1 Like
When i destroy my server and create a new, it’s a clean new install - so all data from the old server is lost
Does anyone know how to get a more detailed error message from the salamander-jupyter notebooks?
I think you already fixed this but it seems errors like this are from torch not compiling properly. I had a similar error from upgrading torch which did not go well (1.01 -> 1.20 = from torch._C import * error).
The fix for my error and possibly a good first thing to try for torch related errors, is to just uninstall torch (instead of upgrading) and then reinstalling latest version.
edit - based on Jeremy’s post, Salamander runs on Conda so pip is not recommended.
Thus:
conda uninstall pytorch
conda install -c pytorch pytorch
That should help ensure proper dependencies are updated.
I am having the same problem, a new server works for a bit, I shut it down, then if I come back to it after a few hours I get the same 500 Internal Server Error. I can still assess the server through terminal, and can assess folders with jupyter, just cannot run notebooks
1 Like
That’s odd. Did you do a conda update or something? Shutting down and starting again shouldn’t be able to cause a server error.
I just shut the server down, went home from work - started it up again, and it’s not working.
Thanks for letting me know.
I have the same problems …once I ve restarted server today I get " 500 : Internal Server Error" from Jupyter.
I ve started a new server today and it auto closed by itself …after starting it again i get the same Jupyter error.
I noticed my code run slowly (on a machine with several GPUs)
I run this line
s = tf.Session(config=tf.ConfigProto(log_device_placement=True))
s.list_devices()
and got
[_DeviceAttributes(/job:localhost/replica:0/task:0/device:CPU:0, CPU, 268435456)]
When I opened another notebook and run the same code all GPUs where identified.
Can it be that tensorflow suddenly cannot see the installed GPUs?
If anyone is still getting this error, and you don’t mind me logging into your instance to debug it, can you please PM me your login username for Salamander so I can take a look?
1 Like