StableSR icon indicating copy to clipboard operation
StableSR copied to clipboard

Why I get killed?

Open misi0202 opened this issue 2 years ago • 8 comments

I use CLI python main.py --train --base configs/stableSRNew/v2-finetune_text_T_512.yaml --gpus GPU_ID, --name NAME --scale_lr False and i get killed 截屏2023-07-07 02 05 08 At first ,i reduce the batch_size and num_workers to 1,but it cant work As a possibility,I use ckpthttps://huggingface.co/stabilityai/stable-diffusion-x4-upscaler I am not sure about that,HELP!!!!!

misi0202 avatar Jul 07 '23 09:07 misi0202

I use CLI python main.py --train --base configs/stableSRNew/v2-finetune_text_T_512.yaml --gpus GPU_ID, --name NAME --scale_lr False and i get killed 截屏2023-07-07 02 05 08 At first ,i reduce the batch_size and num_workers to 1,but it cant work As a possibility,I use ckpthttps://huggingface.co/stabilityai/stable-diffusion-x4-upscaler I am not sure about that,HELP!!!!!

I am back,it is not reachable in ckpthttps://huggingface.co/stabilityai/stable-diffusion-2-1-base,this confused about me.

misi0202 avatar Jul 07 '23 09:07 misi0202

The most possible reason is that your cpu memory is not enough.

IceClear avatar Jul 07 '23 09:07 IceClear

@IceClear Thanks for your reply!! maybe is this problem,but if i cant change me cpu memory,is there have alternative choice?

misi0202 avatar Jul 07 '23 10:07 misi0202

Sry. Currently I have no ideas about which part leads to high memory cost. So no solution from my side.

IceClear avatar Jul 07 '23 11:07 IceClear

Maybe you can check if the model ckpt is loaded twice since in my current setting, the diffusion model and vqgan are initialized from the same large ckpt and I suspect maybe two copies of this large ckpt exist in cpu ram. Not sure about that.

IceClear avatar Jul 07 '23 11:07 IceClear

Maybe you can check if the model ckpt is loaded twice since in my current setting, the diffusion model and vqgan are initialized from the same large ckpt and I suspect maybe two copies of this large ckpt exist in cpu ram. Not sure about that.

Thanks for you patient reply! I will check about that soon!

misi0202 avatar Jul 07 '23 12:07 misi0202

Maybe you can check if the model ckpt is loaded twice since in my current setting, the diffusion model and vqgan are initialized from the same large ckpt and I suspect maybe two copies of this large ckpt exist in cpu ram. Not sure about that.

Thanks for you patient reply! I will check about that soon!

Looking forward to your reply.

IceClear avatar Jul 07 '23 12:07 IceClear

Maybe you can check if the model ckpt is loaded twice since in my current setting, the diffusion model and vqgan are initialized from the same large ckpt and I suspect maybe two copies of this large ckpt exist in cpu ram. Not sure about that.

Thanks for you patient reply! I will check about that soon!

Looking forward to your reply.

SAD,it doesn't work,but thanks to your contribution for this issue,hope I can use it in the future! Have a nice Day!

misi0202 avatar Jul 10 '23 02:07 misi0202