PixelNinjaArt_LoRA_Flux1 Repo update. Uploading the following: I continued training from the original version, saved a 2000 steps and trained. until it was at 4000 steps. I felt the initial version didn`t quite capture the style. So I thought further training was necessary. Turns out my hunch was right. and it looks more accurate to the original style. The LoRA Attached to this commit will be the 4000 step version, but the 2000 step version will be contained in previous. commits. should I or any one else want or need access to that version. Like the last commit, it also includes. an iterations zip file containing all LoRA Iterations produced in this training session( steps 2000 - 4000). That archive also contains a README.txt file Clarifying that the `PixelNinjaArt-Flux1.safetensors` file Contained within is the one trained at four thousand steps. The optimizer.pt archive file has also been updated, So extract that if you want the updated optimizer file.
PixelNinjaArt_LoRA_Flux1 Repo update. Uploading the following: sample images, latents cache archive (.zip), config.yaml (settings used), all LoRA versions packed up in a zip archive, and the archived optimizer.zip
PixelNinjaArt_LoRA_Flux1 Repo update. Uploading the following: sample images, latents cache archive (.zip), config.yaml (settings used), all LoRA versions packed up in a zip archive, and the archived optimizer.zip
