| message_id,timestamp,channel,server_id,author,author_id,content,attachments,attachment_ids,embeds,reactions | |
| 1293254106969800715,2024-10-08T16:50:06.601000+00:00,1053357683417296966,1010980909568245801,vrane4580,251750353458954246,*Tosses away his update button for Forge*,[],[],0,[] | |
| 1293254106969800715,2024-10-08T16:50:06.601000+00:00,1053357683417296966,1010980909568245801,vrane4580,251750353458954246,*Tosses away his update button for Forge*,[],[],0,[] | |
| 1293245623671455798,2024-10-08T16:16:24.025000+00:00,1053357683417296966,1010980909568245801,sergiosantos,225929859153657857,,[],[],0,[] | |
| 1293245623671455798,2024-10-08T16:16:24.025000+00:00,1053357683417296966,1010980909568245801,sergiosantos,225929859153657857,,[],[],0,[] | |
| 1293245515097444362,2024-10-08T16:15:58.139000+00:00,1053357683417296966,1010980909568245801,dodgydude,1086438439056052354,Amazing and comprehensive thank you!!!,[],[],0,[] | |
| 1293245515097444362,2024-10-08T16:15:58.139000+00:00,1053357683417296966,1010980909568245801,dodgydude,1086438439056052354,Amazing and comprehensive thank you!!!,[],[],0,[] | |
| 1293225811696156753,2024-10-08T14:57:40.482000+00:00,1053357683417296966,1010980909568245801,sergiosantos,225929859153657857,"It seems that after the latest Forge update, many users have been experiencing similar issues, especially regarding memory spikes and image generation failures at the last moment. Based on the details of your problem, here are a few possible solutions that might help: | |
| VRAM Offloading: While the --medvram argument has been deprecated, Forge now has a replacement method for VRAM offloading. Make sure you're using the updated argument, such as --vram-offload, which manages memory better for systems with lower VRAM. This should prevent the spikes during the final stages of image generation | |
| VAE and Sampling: If the memory spike coincides with a corrupted output (like the negative or black images with dots), it could be related to the VAE (Variational Autoencoder) or sampling method settings. Try using different VAE settings or reverting to a more stable checkpoint. Additionally, check if your samplers (like Euler or DPM++) are up-to-date and compatible with your models | |
| Flux Realistic Sampler: If you're using the new Flux Realistic sampler, ensure that the associated parameters like ""GPU Weight"" aren't set too high, as this can lead to unexpected behavior, including memory overuse | |
| Check for Extensions: If you have any extensions or additional modules installed, try disabling them temporarily to see if they are contributing to the memory spike and image corruption issue. Some users have reported that after disabling certain extensions, the issue resolves itself | |
| If none of these suggestions solve the issue, you may want to roll back to an earlier version of Forge that worked for you or look into further community discussions on the Forge GitHub, as these problems are actively being addressed in updates.",[],[],0,[] | |
| 1293225811696156753,2024-10-08T14:57:40.482000+00:00,1053357683417296966,1010980909568245801,sergiosantos,225929859153657857,"It seems that after the latest Forge update, many users have been experiencing similar issues, especially regarding memory spikes and image generation failures at the last moment. Based on the details of your problem, here are a few possible solutions that might help: | |
| VRAM Offloading: While the --medvram argument has been deprecated, Forge now has a replacement method for VRAM offloading. Make sure you're using the updated argument, such as --vram-offload, which manages memory better for systems with lower VRAM. This should prevent the spikes during the final stages of image generation | |
| VAE and Sampling: If the memory spike coincides with a corrupted output (like the negative or black images with dots), it could be related to the VAE (Variational Autoencoder) or sampling method settings. Try using different VAE settings or reverting to a more stable checkpoint. Additionally, check if your samplers (like Euler or DPM++) are up-to-date and compatible with your models | |
| Flux Realistic Sampler: If you're using the new Flux Realistic sampler, ensure that the associated parameters like ""GPU Weight"" aren't set too high, as this can lead to unexpected behavior, including memory overuse | |
| Check for Extensions: If you have any extensions or additional modules installed, try disabling them temporarily to see if they are contributing to the memory spike and image corruption issue. Some users have reported that after disabling certain extensions, the issue resolves itself | |
| If none of these suggestions solve the issue, you may want to roll back to an earlier version of Forge that worked for you or look into further community discussions on the Forge GitHub, as these problems are actively being addressed in updates.",[],[],0,[] | |
| 1293223103400513558,2024-10-08T14:46:54.774000+00:00,1053357683417296966,1010980909568245801,dodgydude,1086438439056052354,"Guys... its flippin happening again... my Forge webui is generating fine and then as the image is saving (literally gen is a 98%) and then as it hits 100% my memory spikes (which was fine the whole way through) and it bombs out a reg and green negative looking version of the image with black dots on it. This happened after a recent forge update. Previously I could insert --medvram into the launch arguments, now thats been done away with. I've tried the suggested offload vram argument suggested by Forge to mimic the conditions of --medvram, but its still not working. Any suggestions would be greatly appreciated! Aaaaaah!",[],[],0,[] | |
| 1293223103400513558,2024-10-08T14:46:54.774000+00:00,1053357683417296966,1010980909568245801,dodgydude,1086438439056052354,"Guys... its flippin happening again... my Forge webui is generating fine and then as the image is saving (literally gen is a 98%) and then as it hits 100% my memory spikes (which was fine the whole way through) and it bombs out a reg and green negative looking version of the image with black dots on it. This happened after a recent forge update. Previously I could insert --medvram into the launch arguments, now thats been done away with. I've tried the suggested offload vram argument suggested by Forge to mimic the conditions of --medvram, but its still not working. Any suggestions would be greatly appreciated! Aaaaaah!",[],[],0,[] | |
| 1293176342979805265,2024-10-08T11:41:06.221000+00:00,1053357683417296966,1010980909568245801,vrane4580,251750353458954246,I mean... It technically is a failed gen xD,[],[],0,"['🤔:1', '😆:1']" | |
| 1293176342979805265,2024-10-08T11:41:06.221000+00:00,1053357683417296966,1010980909568245801,vrane4580,251750353458954246,I mean... It technically is a failed gen xD,[],[],0,"['🤔:1', '😆:1']" | |