jp-Elle-2325 is the latest in my attempt to learn TI's. <ul><li>Elle is not her real name.</li><li>I made this embedding with 30 images I took of the model from a couple of photoshoots I did years ago. There is about 2 years age difference between when the shoots took place and I was still learning (I still am of course) photography, When you don't specify a background or location you'll usually get a sheet or muslin in background as most of the images have the same backdrop. I did a decent job in the text files pulling it out so she does well with a variety of backgrounds.</li><li>While this model is close it does not reproduce any of the training images exactly. I may do another version.</li><li>30 images trained for 2325 steps.</li><li>16 vectors per token.</li><li>I used the following progressive learning rate 0.05:10, 0.02:20, 0.01:60, 0.005:200, 0.002:500, 0.001:3000, 0.0005</li><li>Batch size of 1 with 1 gradient accumulation step.</li></ul>I own the images used as I took them so I'm going to be much more liberal with permissions and commercial use.As always I appreciate your feedback and look forward to seeing what you all create.

Files with SHA256: 7003f89f63b776d5a3666a7c50f8d520e7038f51b6a3a0130cd9c076e8e4e065