Intro

I am using Adele Exarchopoulos for a lot of my training mostly because I have high respect for her acting and she happens to be easy on the eyes to work with. A lot of images were taken from her role in Blue is the Warmest Color in my first training so I also added some non-movie images to increase variety.

Pre Training

I gathered 70 images of Adele Exarchopoulos. I am using both facial images (from previous training) as well as normal photos. None of the images were full body as they were still very close up. These are 512x512 images instead of 1024x1024 images because I don’t have the specs to train a 1024x1024 model. I used Blip captioning to generate the filewords and edited each individually to reduce potential hallucinations.

Training

I used 0.005:100:0.0025:250,0.001:500,0.0005:1000,0.00025 for my learning rate. I am going for 10K training steps total. I am using a batch size of 5 with Gradient Accumulation Steps set to 3. The estimated time of completion is 6 hours. For the embedding I am using 8 vectors per token.