search

tencent-ailab / Frequency_Aug_VAE_MoESR

Latent-based SR using MoE and frequency augmented VAE decoder
Apache License 2.0
149 stars 4 forks source link

请问 8x SR 需要多大的显存？目前 16G 的卡会 OOM #10

Open shengabcqwe opened 1 year ago

YUANMU227 commented 1 year ago

same problem how to fix this problem

viperyl commented 1 year ago

so far, i just load all of the model, the vram consumed 22Gb

rfan-debug commented 11 months ago

How did you manage to launch it with 22GB vram consumption? I tried it with 24GB GPU but ran into OOM

include5636 commented 7 months ago

How did you manage to launch it with 22GB vram consumption? I tried it with 24GB GPU but ran into OOM

Have you solved this problem? I ran into OOM too by using 24GB GPU

viperyl commented 5 months ago

i use this model for daily upscaling work, and VRAM cost is around 27.47 GB, i will considering optimize this model later