Training flux-schnell model

How do you manage to train a flux-schnell model using serverless, i have loaded the images using s3 Bucket, but what about the waiting time of the training process ? wont i get a timeout while waiting 20 min for the training process to end?
9 Replies
yhlong00000
yhlong000004mo ago
serverless is design for inference, it's simpler to just use a pod.
Untrack4d
Untrack4dOP4mo ago
the thing is i want the scale up feature and pay only for usage
yhlong00000
yhlong000004mo ago
Are you planning to train the model or just run inference? You can start and stop the pod to save on costs as well.
Untrack4d
Untrack4dOP4mo ago
I want to do both, but the inference part its easy, the training one its tricky, i tested the training using pod, the thing is i now want it to scale up when i want to make more than one training at once And if i stop, when i start it again i may have lost my GPU, thats happeing all the time, and its a pain in the ass
yhlong00000
yhlong000004mo ago
try to use the network storage and save your stuff there, that way you won't tie to a specific machine.
Untrack4d
Untrack4dOP4mo ago
but i still want the ability to scale autonomously have you manage to use the network storage?, i cant really find any documentation on that (for thye serverless)
pecet
pecet4mo ago
hi, how you train flux on pod? is there any docker image? I meann I am looking for no gui solution
Untrack4d
Untrack4dOP4mo ago
Me to, I am using ai-toolkit
Encyrption
Encyrption4mo ago
With serverless you do not have access to many of the higher level GPU like you do with pods. Serverless was designed with inference in mind, not training. Not saying you cannot train on serverless but you are blazing your own path in doing so.
Want results from more Discord servers?
Add your server