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
5-11 stable horde announced a 25% reduction for worker rewards. But they worded it as only affecting the (current) official worker.
Their worker depends on Nvidia (CUDA toolkit) support which will never happen on Apple M series. So this kind of sucks. I no nothing of the alleged new worker. Probably the same.
The full announcement, and I copy-paste:
2023-05-11 With the upcoming deployment of the hordelib-based workerJug has looked into creating a more efficient model to determine generation kudosinstead of reusing the numbers I hallucinated one day.He used what we know best and we trained an explicit model to calculate kudos, based on the performance of his own GPU on the comfy branchThis new calculation should be much more accurate in terms of things like controlnet and resolution impact.The good news is that the new comfy branch this seems to reduce kudos costs for high resolutions accross the board.Note: Due to the current worker (based on nataili) being slightly lower quality at the benefit of speed, and thus getting a boost due to the new kudos modelWe have implemented a 25% reduction for its rewards to bring it up to line with its actual performance.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
5-11 stable horde announced a 25% reduction for worker rewards. But they worded it as only affecting the (current) official worker.
Their worker depends on Nvidia (CUDA toolkit) support which will never happen on Apple M series. So this kind of sucks. I no nothing of the alleged new worker. Probably the same.
The full announcement, and I copy-paste:
2023-05-11
With the upcoming deployment of the hordelib-based workerJug has looked into creating a more efficient model to determine generation kudosinstead of reusing the numbers I hallucinated one day.He used what we know best and we trained an explicit model to calculate kudos, based on the performance of his own GPU on the comfy branchThis new calculation should be much more accurate in terms of things like controlnet and resolution impact.The good news is that the new comfy branch this seems to reduce kudos costs for high resolutions accross the board.Note: Due to the current worker (based on nataili) being slightly lower quality at the benefit of speed, and thus getting a boost due to the new kudos modelWe have implemented a 25% reduction for its rewards to bring it up to line with its actual performance.
Beta Was this translation helpful? Give feedback.
All reactions