Are we being billed for data in the ".Trash" folder?

What application are you running that creates/uses a .Trash directory in your persistent storage filesystem—rather than, for example, in your home directory?
Edit: JupyterLab?
Edit 2: Noting that the Trash Specification allows for .Trash directories in the “top directory” of filesystems, like Lamb…

Does Inference API support batch/asynchronous processing

@venkatasg The Lambda Inference API currently doesn’t support batching.

Model and content limit

@Masira There are no limits:

The Lambda Inference API enables you to use large language models (LLMs) without the need to set up a server. No limits are placed on the rate of requests. The Lambda Inference API can be used as a drop-in replacement for applications currently using the OpenAPI API. …

Inference API Limits?

@junruilee There are no rate limits.
I just now added to the docs, “No limits are placed on the rate of requests.”

Any 80+ GB VRAM GPUs coming to european data centers?

Hi @AndreasBe, please contact Lambda’s Sales team to learn about capacity in EU data centers. Be sure to reference this thread.

Hermes 405 Issues

@Jason I was told the weights haven’t been touched.

Inference API call returning "HTTP/1.1 400 Bad Request"

Can you try the Using the Lambda Inference API > Creating chat completions cURL example and post the results?

User Generate Content

Additionally, regarding data collected from usage of the service as per the privacy policy, does this include both the input and output from Lambda Chat as well?

Input and output are encrypted in transit. Sorry, this is with respect to the Inference API.
With respect to Lambda Chat, the chat …

Why is 1x H100 SXM not available

@minemaster01 Lambda has them but they’re all taken.

GH200 not working with attached file system

@doctorjuice Are you still encountering this issue?

PyTorch DDP NCCL hangs on h100 server

I wasn’t able to replicate the issue.
Can you provide step-by-step instructions to replicate the issue?

[request] please add deepseek to lambda inference

This is being worked on and should be generally available in a couple of weeks.
FWIW, the models can already be tested on Lambda Chat.

Lambda Labs Cloud Service Showing White Screen Help Needed

@tekidap Can you provide the logs from your browser’s dev tools?

Cheapest way to upload data into the filesystem with slow upload bandwidth

@dushan Unfortunately, you’re correct that at this time, you need a GPU instance to use a filesystem. It’s a bit painful, but you can upload your data somewhere (to Cloudflare R2, for example) then download your data from there directly to your instance.

Unable to enable mig on A100 instance

MIG settings don’t persist across reboots.
But, take a look at MIG Partition Editor, specifically, nvidia-mig-manager.service.

SSH connection is stuck

@Tushar If you haven’t already, please submit a support ticket → Lambda Support

Lambda Labs <-> Azure VPC

Have you looked into VPN solutions? Tailscale is easy to set up. Also look at traditional VPN solutions like WireGuard and OpenVPN.

Serve model endpoint api via public url

So, what is the best way to expose the model api to public but with only HTTP traffic allowed in firewall?

I suggest using a reverse proxy. I’m happy to help with setting this up. See my gradio-load-balancing project for inspiration → GitHub - cbrownstein-lambda/gradio-load-balancing: Proof of co…

Lambda <> Openrouter Woes

@Jason @friendly_fox Small update but it’s all I have at the moment: Lambda’s engineers are slowly easing the rate limiting for the Inference product. I’m trying to get an ETA for when the rate limiting should stop affecting your projects.

Is GPUDirect Storage supported for A100 and H100 on demand cloud?

GPUDirect Storage isn’t available.

Package Update Recommendations (Mellanox , Lambdalabs)

sudo apt update && sudo apt upgrade is recommended.
If you want to be extra cautious, back up your data before upgrading, but it’s rare that the above command causes problems.

Unable to boot into AI Desktop workstations

Install Ubuntu Desktop 24.04 → https://ubuntu.com/download/desktop
Install Lambda Stack → Lambda Stack: an AI software stack that's always up-to-date

How can I maximize speed on Lambda GPU instances?

Can you specify in greater detail what you mean by the training is slow? If you have log files, that might help with troubleshooting the issue.

Is Lambda Stack free for use on non-LambdaLabs computers

Anyone is free to use Lambda Stack.
This Lambda GitHub repo might be helpful → GitHub - lambdal/lambda-stack-dockerfiles

Ubuntu 24.04 support

Ubuntu 24.04 is validated and supported.

Lambda Chat Completions API Cost

The API is free to use at the moment but your Cloud account needs to have a card on file. You won’t be billed any amount unless you launch an instance.

Enable full disk encryption

The easiest way to enable full disk encryption on a new system is to use the recovery image. At the disk setup step, choose Erase disk and install Ubuntu > Advanced features > Use LVM and encryption.

Cannot run jupyter notebooks, installed modules are not found

Hi,
Lambda’s engineers are working on a permanent fix.
For a temporary workaround, please run the following command:

echo "c.InteractiveShellApp.exec_lines = ['import sys; sys.path.append("/home/ubuntu/.lambda/lib/python3.10/site-packages")']" > ~/.ipython/profile_default/ipython_config.py

Issue with pandas read_excel() function

Hi,
Lambda’s engineers are working on a permanent fix.
For a temporary workaround, please run the previous command.

Downgrading Cuda 12.4 to 12.1

Lambda usually recommends to customers that they use conda virtual environments to use different CUDA versions than what’s preinstalled.
See also: Cuda | Anaconda.org

Power Supply Failure

I’d submit a support ticket if you haven’t already → https://support.lambdalabs.com/hc/en-us/requests/new

No direct startup script support?

Check out Lambda’s guides on:

Deploying models with dstack Using SkyPilot to deploy a Kubernetes cluster

Where can I find pricing?

See: GPU Cloud - VMs for Deep Learning | Lambda

Is there a way to export usage for billing purpose on a daily basis?

There currently isn’t a way to do this.

Bypassing lambda stack when installing libraries with pip inside conda env

Have you tried running conda install pytorch torchvision instead of installing with pip?

Cannot lauch new instance

I believe new launches were paused for maintenance, but you should be able to launch instances now.