Runpod

R

Runpod

We're a community of enthusiasts, engineers, and enterprises, all sharing insights on AI, Machine Learning and GPUs!

Join

⚡|serverless

⛅|pods

🔧|api-opensource

📡|instant-clusters

🗂|hub

ComfyUI template extra path bug

The official template isn't picking up the workspace folders for ComfyUI because the yaml extension is wrong. Here's the fix. https://github.com/runpod/containers/pull/92...

Network Volume Utilization

I'm using RunPod to train my model. And I use RunPod network volume to store my artifacts. I created a volume with 60GB, and it's path is /workspace. When I finish uploading my data, it says the network volume is already at 80% utilization. While when I do du -sh /workspace/, it only shows 17GB, (which is what I expected), but why I'm seeing 80% utilization, do I need to do some cleaning, after unzip my data? The volume is on EU-RO-1...

createing a pod in japan through sdk

I'm trying to create a pod using the following configuration: pod = runpod.create_pod( name="my-h100-pod",...

How in the hell do I get my files off my pod... ima pull my hair out...

How do my files off of the damn pod... this is the dumbest design I ahve ever seen. It almost makes sense and then is like of die in hell if you want your files... complete trash from a users standpoint... I dont mind paying but what the actual fuck

Being charged for inaccessible pod

When I create a pod, I start getting charged even if it's not yet accessible. I was told that charges only start once the pod is fully connectable, but that's clearly not the case now. Why is everything so inconsistent and misleading? Is this a bug, or is this how it's actually supposed to work?

Having issues connecting on desktop

Hey guys, so I'm trying to connect to run pod on my desktop computer and it does not load up the main screen. Just cycles the loading screen. When I do connect all of the pods are unavailable and the templates are limited. When I try this on my Samsung phone I connect immediately and all of the pods templates including the one I need are there. ...
Solution:
it was. thank you

Site-to-Site VPN

Is it possible to setup a Site-to-Site vpn with a POD to secure the traffic between our Pod and our internal organizational network? Unfortunately I could not find any information about this. I am aware of your global networking features and end-to-end encryption, but this would involve a direct link between our infrastructure and the RunPod environment.

Official template fail to start

runpod/pytorch:2.8.0-py3.11-cuda12.8.1-cudnn-devel-ubuntu22.04 GPU: 4090 create container runpod/pytorch:2.8.0-py3.11-cuda12.8.1-cudnn-devel-ubuntu22.04...

Python venv setup in my network storage doesn't work after spinning up another pod

I setup a python virtuasl environment on my network storage and it was working perfectly with my dual A6000 pod. Now, I've spun up a dual 4090 pod and the virtual environment cannot locate modules when running my python script. All the modules are present in the network storage folders where they should be.

Cannot connect to web terminal with dual 4090s

I just spun up a pod with 2 4090s and it won't let me connect to the web terminal. I click 'Start', it says 'Starting' for a second and then goes back to 'Stopped'. This is not an issue when I spun up a dual A6000 pod.

POD deleted despite Network volume

We have critical code running on Runpod that we stored in a VM volume. This got deleted, alongside the volume, despite us paying the bill a few days ago. We urgently need the code back

Trying to Connect to Pod Storage with SCP on WinSCP

couple of months ago I was able to connect to my pods to transfer files for my comfyui. I made new keys today and updated the public key in my use settings for run pod and linked the private key file for ssh tunnel in my winscp connection settings but have no luck. I'm getting connection refused respons trying to ssh to the address from terminal as well but some how connecting to ssh without a direct TCP connection is working. I had it working a couple of months ago with WinSCP but no longer can get it connected....

How do I transfer files into my network volume? Is that the /workspace folder on my pod?

I'm unclear on how to utilize the network volume attached to my pod. Is the /workspace folder connected to the network volume? I have a 120gb network volume attached and barely installed anything into the workspace folder and it's already at 6% full. Any guidance would be appreciated. Thanks.

Template issues

1. I'm aware these are community templates, so let's be real i know i'm not going to get refunds LOL 2. I'm just passing the frustration along as i'm not sure there's any real support for these... 3. Sadly i haven't used runpod in al ong time and other services work differently (am not complaining, just ... haven't really been on your service in a while) ok so.....
Solution:
Update: Swapped to forge regular and the models are working, inclusive of the ones that are merged.. I copied everything over from the old forge install - sincei t's network volume.. And suddenly BANG it works. I think there's some weird thing where a code mixes itself up-- Like i dont know whats weirder .....

Is it possible to do GPU profiling on Runpod's pods

I want to rent a GPU for profiling on Runpod with ncu. I tried to do so but got the permission issue as detailed in this link: https://developer.nvidia.com/nvidia-development-tools-solutions-err_nvgpuctrperm-permission-issue-performance-counters

Global Networking - Official Template - Not Working

Hello, I created two pods using the official Pytorch 2.8.0 template, each with "Global Networking" enabled, then tried to have the nodes ping each other aligning with what's described in the documentation: https://docs.runpod.io/pods/networking ```...
No description

Creating a pod using a Docker image built via Serverless GitHub integration

I'd like to create a pod on Runpod using the docker image I built via GitHub integration for the Serverless endpoint. I can't build it locally and push it to my registry because then the container doesn't work properly. It only works when it is built on the Runpod platform via the GitHub integration. In the build logs I can see it is published in a registry but I can't see its address. I assume it's a Runpod's registry so it'd probably have access to it during pod creation?...

Slow Model Loading - Solutions

Problem Description Loading HiDream-I1-Fast and Llama-3.1-8B-Instruct on RunPod (A100) takes ~34s (total ~46s) with only 13% CPU usage. Profiler shows aten::copy_ (71% CPU, 21.69s) and cudaMemcpyAsync (24%, 7.42s) as bottlenecks. (I checked other API providers — they offer this model via API at a much lower cost than what I'm currently testing, and they deliver outputs in under 15 seconds idk how!), but I couldn't optimize loading time despite trying multiprocessing with spawn (caused semaphore leaks). Need to reduce loading time as much as possible. Any solutions or insights?...

all installed py libs are consistent except yfinance

Hi as topic says, i have several py libs installed and all are consistent after stopping and starting the POD except yfinance, is there sth special about yfinance ?

Unable to download huggingface model in runpod . facing issue : OSError

Hi , i am not able to download huggingface model in runpod . facing issue OSError: unsloth/gemma-3-27b-it does not appear to have files named ('model-00005-of-00012.safetensors', 'model-00006-of-00012.safetensors', 'model-00007-of-00012.safetensors', 'model-00008-of-00012.safetensors', 'model-00009-of-00012.safetensors', 'model-00010-of-00012.safetensors', 'model-00011-of-00012.safetensors', 'model-00012-of-00012.safetensors'). Checkout 'https://huggingface.co/unsloth/gemma-3-27b-it/tree/main'for available files. can you anyone please help...
No description