You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When using .enable_model_cpu_offload() if the model is currently offloaded to cpu, generating a normal image waits for the model to load back onto the gpu before generating. When using hypertile, it does not wait for the model and I get found at least two devices, cuda:0 and cpu! on the first generation. It works fine after that until offloading occurs and then once again the first generation will fail.