Update README.md

README.md

@@ -21,6 +21,8 @@ source_datasets:
 
 # E621 2024 SDXL VAE latents in 1k tar #
 
+- Dedicated dataset to align both [NebulaeWis/e621-2024-webp-4Mpixel](https://huggingface.co/datasets/NebulaeWis/e621-2024-webp-4Mpixel) and [deepghs/e621_newest-webp-4Mpixel](https://huggingface.co/datasets/deepghs/e621_newest-webp-4Mpixel). "4MP-Focus" for average raw image resolution. 
+- Latents are ARB with maximum size of 1024x1024 as the recommended setting in kohyas. Major reason is to make sure I can finetune with RTX 3090. *VRAM usage will raise drastically after 1024.*
 - Generated from [prepare_buckets_latents_v2.py](https://github.com/6DammK9/nai-anime-pure-negative-prompt/blob/main/ch06/sd-scripts-runtime/prepare_buckets_latents_v2.py), modified from [prepare_buckets_latents.py](https://github.com/kohya-ss/sd-scripts/blob/sd3/finetune/prepare_buckets_latents.py).
 - Used for [kohya-ss/sd-scripts](https://github.com/kohya-ss/sd-scripts/blob/sd3/docs/train_README-ja.md#latents%E3%81%AE%E4%BA%8B%E5%89%8D%E5%8F%96%E5%BE%97). In theory it may replace `*.webp` and `*.txt` along with [meta_lat.json](https://huggingface.co/datasets/6DammK9/e621_2024-latents-sdxl-1ktar/blob/main/meta_lat.tar.gz).
 - It took me around 10 days with 4x RTX 3090 to generate (with many PSU trips and I/O deadlocks). Perfect case would be 3-4 days only (18 it/s).