R
RunPod•4w ago
Orazur

Fast loading of large docker image

Hello, I am trying to use a large docker image (>20GB) to start a pod. Is there a way to cache it in a network volume, and then start a pod from it, to start a pod quickly (I already sent my image in .tar on a volume, but couldn't find how to start a pod from it)? Or is there a better solution? Thank you!
2 Replies
justin
justin•4w ago
You move anything that is of large assets to a Network volume, I recommend to do this in a region that has both CPU Pods and GPU Pods availiable in case you ever need to spin up a CPU Pod + network volume to move data around. This should drastically decrease your docker image size, and let you spin up containers much faster You don't send your image to a .tar rather the things that make your image so large, and use ur docker image more as the actual run time processing unit
Orazur
Orazur•4w ago
Thank you very much, I’m Going to try that! I tested and it’s working well, thanks again!
Want results from more Discord servers?
Add your server
More Posts
0 gpu availablecurrently i can't assign a gpu to my pod. will i get the option to at a later point when some free uDATA LOSS IN EU-RO-1 - URGENTNeed someone to communicate ASAP this is literally sev0. We have all of our data and work backed in Clarify RAM availableHey there! Was thinking of using 4090 pods, and I saw that the deployment only had 60GB of RAM - whiCreate CPU Pod through GraphQLThe API seems to expect a gpuTypeId even when you specify gpuCount: 0. Is there currently any way ofHow to add files?I just setup a pod for the first time and I am trying to ssh into it to add some custom model files.Very Slow MappingHello! I am trying to run `dataset.map()` and it takes only a few minutes when I run it on Colab. HoCPU Instances on 64 / 128 vCPUs FAILI can deploy my app on all instances except for 64 & 128 vCPU. Both of these run on AMD EPYC 9754 1JS pupeteer issueHi, I have a pupeteer chromium task that i want to run on the runpod which is written in typescriptPytorch Lightening training DDP strategy crashed with no error caught on multi-GPU workerIt looks like serverless worker will crash when spawning new processes from the handler. It crashes How to get a public URL?I don't want localhost because I can't access it locally.