bkrs
/

ICBINF

@@ -16,6 +16,8 @@ tags:
 <img src="https://huggingface.co/sda7/ICBINF/resolve/main/00734-7b6741ddf6-3289842378.png" alt="image" width="640"/>
 Yet another XL shittune, don't expect much from it, it's more like artist knowledge reinforce, with combination of datasets from [here](https://civitai.com/models/723360/ai-styles-dump-illustriousrouweinoob) and [here](https://civitai.com/models/808215/illustriousnoob-vpred-xl-148-styles-collection) it was trained from noob-vpred1.0 checkpoint. List of all styles that was trained can be found [here](https://huggingface.co/sda7/ICBINF/blob/main/artists_v3.228.txt)
 First important thing - it's bf16 precision checkpoint, doesn't really matter at inference time, but for maximum numerical precision launch latest reforge with --unet-in-bf16 --bf16-text-enc keys or comfyui with --bf16-unet --bf16-text-enc keys. But for training it's preferrable to use bf16 with edm2 weights.

 <img src="https://huggingface.co/sda7/ICBINF/resolve/main/00734-7b6741ddf6-3289842378.png" alt="image" width="640"/>
+1.5 version are continuation of training with a few more added styles, [styles list](https://huggingface.co/sda7/ICBINF/resolve/main/v1.5/p4-unet-tenc/artists_v3.228.txt) included in this version.
 Yet another XL shittune, don't expect much from it, it's more like artist knowledge reinforce, with combination of datasets from [here](https://civitai.com/models/723360/ai-styles-dump-illustriousrouweinoob) and [here](https://civitai.com/models/808215/illustriousnoob-vpred-xl-148-styles-collection) it was trained from noob-vpred1.0 checkpoint. List of all styles that was trained can be found [here](https://huggingface.co/sda7/ICBINF/blob/main/artists_v3.228.txt)
 First important thing - it's bf16 precision checkpoint, doesn't really matter at inference time, but for maximum numerical precision launch latest reforge with --unet-in-bf16 --bf16-text-enc keys or comfyui with --bf16-unet --bf16-text-enc keys. But for training it's preferrable to use bf16 with edm2 weights.