Thank you for the information. Yes, I do have the AWS $500 credits.
Jeremy - On Lesson 2-Image-models notebook, it does
os.makedirs('/cache/planet/tmp', exist_ok=True). But there’s permission denied when running the notebook. Should it be /cache or can it be ~/cache?
Looks like you’re doing the ‘crestle only’ bit, but you’re not in crestle…
After connecting to the AMI using ssh and starting Jupyter Notebook, I am unable to connect to the jupyter notebook URL - when I paste it into my browser I get the following:
This site can’t be reached
localhost refused to connect.
Search Google for localhost 8888
I am new at this, but am wondering why it is telling me to connect to localhost:8888 - I would have thought that this would be hosted somewhere at Amazon?
Please refer to the comments in the beginning of this thread and this post regarding SSH tunneling.
I got it working, by appending the following to the ssh connection string - as explained above by @wluo - thanks to him!
Hi! I successfully created an instance with this AMI and launched lesson 1 notebook, but since it is Oregon region I’m getting huge latencies which makes very hard to work. Any way to reduce those or launch an instance in another region ?
Yes I’ll create one in Europe and Asia before class starts.
Yay Looking forward to the EU images.
Also, is it possible to share the base-ami id and scripts/steps you ran on it, maybe checked into the git repo somewhere?
Glad to see it worked for @Chris_Palmer, here is a more detailed step by step guide to run the jupyter notebook with AWS AMI:
To run p1v2 lession 1 notebook on AWS fast.ai AMI:
make sure you have a key pair generated and put your private key .pem file under your linux home dir under .ssh dir. Make sure the key pair is for the region of US Oregon, since the fast.ai AWS AMI, ami-8c4288f4, is for now for US Oregon only. If you have a key pair for other region, you can import it to Oregon by providing the key pair name and the public key, which can be generated by this command below, and when prompted, provide the path to your .pem file
$ ssh-keygen -y
make sure to choose the fast.ai AMI ami-8c4288f4 under the region of US Oregon, and choose a GPU instance such as g2.2xlarge or better, and make sure the IPv4 Public IP is enabled
start your instance, after it’s running, go to your local linux prompt, issue this command:
$ ssh -i <path_to_your_pem_file> ubuntu@<your_instance_public_ip> -L 8888:127.0.0.1:8888
once logged in, do the following to get the most updated code:
$ cd /home/ubuntu/fastai
$ git pull
then go to the course dir and launch the notebook:
$ cd /home/ubuntu/fastai/courses/dl1
$ jupyter notebook
from the console, copy the url below to your browser, no need to replace localhost with your public ip:
double click lesson1.ipynb to launch the notebook for lesson 1 and enjoy …
Please ask if you have further question, or add to this guide if I missed something
I know p3.2x might be a bit of an overkill. What’d be a better choice?
A g2.2x or p2.x?
I just got my EC2 instance limit increased and was successful in opening a Jupiter notebook on the AMI !! Thanks @wluo for the detailed steps.
Thanks for the detailed guideline
The AMI is ami-8c4288f4
p2 is probably what you want.
Would you be able to create one in Australia? There are not many of us in the class but our closest server is in Sydney…
This is the ID of the Sydney AMI:
ami-39ec055b - taken from the Wiki: Lesson 2 thread.
Thanks @reshama! Could you add a link to that to the top of the lesson’s wiki post? It’s really great and I’d like everyone to be able to see it