Using docker context with Lambda VM

I’m not at a machine currently to verify, but you have sudo access on the default “ubuntu” user. Couldn’t you grant access to the Docker group to this user as a one time operation for each instance and then use Docker Context from your laptop normally?

sudo usermod -aG docker ubuntu

Sunsetting Chat == Sunsetting Inference?

I have the same question. The email I received mentions that this includes the “chat API” - was there a separate API for the chat app or will this include the Inference API? I would be quite sad if Lambda’s inference API shuts down. The reliability of the throughout is something I can’t find on other platforms.

Feature Request: Spending Limits

I can imagine how creating usage limits on rented instances could get messy (example: every time a long running instance breaches the spend-limit, the provider has to decide to eat the cost and track a debt or kill what might be a mission-critical non-resumable workflow for someone). With this in mind...

Script to automate listing available instances and starting an instance?

It might be overkill but if you enjoy using Terraform to automate your infrastructure I’m using my own custom Lambda GPU Cloud Terraform Provider for these sorts of things. You can specify “any” as an instance type, create/destroy SSH keys, set up default packages and setup tools/scripts. It’s still...

Automatic Creation Of Virtual Machines

I’ve got a Terraform provider that I use for this type of thing. Lambda’s GPU Cloud API lends itself very nicely to infrastructure as code. I’m going to be cleaning it up and making it public soon and will ping you when that’s ready, but for now check out the latest post. Another user has made a very interesting implementation that...

PyLambdaCloud: simple way to launch tasks from Python

Nice! Excited to try this out when I get some spare cycles later this week.

Upgrading to later version of Python

While I’m not much of a Jupyter user personally, it seems like get_ipython().system() does not respect your current virtual environment in the same way that you may expect. It’s not running a session with the virtual environment “activated” in the same way (meaning $PATH’s are not set to point to your...

Upgrading to later version of Python

What you’re looking for is doable in the Web JupyterLab that Lambda sets up for you. Here is a quick demonstration:

Step 1: Setting up a normal instance

Nothing special. 1x_A10 if curious, but not important for this demonstration:

Step 2: Installing Python3...

Docker data saved in persistent storage

In short, yes you can definitely do this and you have root access to your instances. Less of a headache than symlinking (for me, largely preference) is just telling Docker that you’ve got your own preferred data-root location:

I’ve just run through this myself on two Lambda Cloud GPU Instances just...

Too many open files

The next time this happens can you capture the output of:

# print [num-handlers] [PID] for top 10 processes
lsof -n | awk '{print $2}' | sort | uniq -c | sort -rn | head -n 10

then the names of the processes attached to each PID (should be the second arg) with:

ps aux | grep -i <PID>

LambdStack on WSL2 vs ACPI

I don’t have a setup to perfectly mimic this, so apologies in advance at what is mostly a guesswork answer. Trying to recreate on my only Windows machine did not encounter any issues, however without realizing I’d installed it on WSL1 and only LATER upgraded to WSL2 - so perhaps that’s worth a shot...

Import Environment YAML to Cloud GPU via conda

There are a number of possible reasons that you’d want to be doing this so I’ll try and cover an example for how to tackle each of them:

Scenario 1: Install common packages on the Lambda GPU Instance

This one is very straightforward. On your local machine...