Sign In

Runaway LoRA training images, pt. 1

Facial_Deconstruction_20231224161529_e000002_02.png
Facial_Deconstruction_20231224165104_e000006_01.png
Facial_Deconstruction_20231224182103_e000002_00.png
Facial_Deconstruction_20231224183325_e000003_01.png
Facial_Deconstruction_20231224184505_e000004_00.png
Facial_Deconstruction_20231224185713_e000005_01.png
Facial_Deconstruction_20231224190840_e000006_00.png
Facial_Deconstruction_20231224192048_e000007_01.png
Facial_Deconstruction_20231224193247_e000008_00.png
Facial_Deconstruction_20231224193300_e000008_01.png
Facial_Deconstruction_20231224195658_e000010_01.png
Facial_Deconstruction_20231224203350_e000013_01.png
Facial_Deconstruction_20231224203404_e000013_02.png
Facial_Deconstruction_20231224204558_e000014_01 (1).png
Facial_Deconstruction_20231224204611_e000014_02.png
Facial_Deconstruction_20231224204623_e000004_01.png
Facial_Deconstruction_20231224205818_e000015_00.png
Facial_Deconstruction_20231224205832_e000015_01.png
Facial_Deconstruction_20231224205846_e000015_02.png
Facial_Deconstruction_20231224210916_e000006_01.png

I've been working on a collage based LoRA recently, but my latest training attempt was affected by some sort of computer error resulting in repeating failures and simultaneous resubmissions. The LoRA would be training on epoch 1 and epoch 11 (then 2 and 12, etc.) at the same time, my screen occasionally flipping back and forth between the two!

I discovered a typo in one of the tags, so I wouldn't save any of the models, but I did save example images I liked, which I share with you now.