Any hacky way to increase payload size?
Can i send it in chunks or something? i need serverless to process uploaded video.
RSend a link and have your worker download it
Pthat would work, but implies that i also deploy some storage service, do you recommend something?
personal google drive, runpod network volume?
clean up strategy?
personal google drive, runpod network volume?
clean up strategy?
RI assumed you already had that since you mentioned uploaded video.
For personal use google drive could work yea. If it's for a automated processes , then any s3 type of storage solutions could work and you'd probably need api endpoints so your runpod workers can communicate with your server
For personal use google drive could work yea. If it's for a automated processes , then any s3 type of storage solutions could work and you'd probably need api endpoints so your runpod workers can communicate with your server
YYou might want to use cpu serverless for uploading to network volume, then ask GPU serverless to processing it. It will more cost effective
CBut are CPUs available in CPU Serverless?
C
Pyeah, ok i got it, i thought i could do it lazy way, just send base64 encoded video directly to run endpoint(( making a full featured service is a bit overwhelming
DWhen you send a job for serverless, you are sending a queued job with a payload. This is why there’s size limits.
Vfinally just ditched base64 and rent storj now, you upload videos there then either get a signed link or pass the file key, server processes it and uploads the response, had to tweak my worker-comfyui version a bit to make it work with more than just images
F@PBG you can use new load balancer serverless type, that doesn't have the same size limit
CNice idea. Have recently started learning Serverless. I have a Mega subscription, so Mega S4 so should really help me achieve the similar stuff without needing to rent storage.
Kif you are comfortable limiting to a single datacenter, you could also utilize a network storage volume with S3 access. Simply upload the data to the network volume via the S3 api, and then in the request send the path and the worker should be able to simply open the file as if it was on the filesystem
