It's started working slowly, though the showcase images with seed are coming up very different. Still, better than the auto fails of before!

Back to breaking even with absolutely zero loras and embeddings on the civit generator :( a damn shame as its my SD go to before I try others. Especially as it includes photon, which I love for realism, which you can't use otherwise.

v3 not working, errors everywhere :)

Things based on diffusers don't seem to like it (v2 is fine). I needed to use the following python script to rewrite the .safetensors file before it would work with what I usually use (though its fine in ComfyUI):

(No idea whether civitai will format this correctly (apparently not))

```python
import sys

import os

from safetensors import safe_open

from safetensors.torch import save_file, load_file

def fix_diffusers_model_conversion(load_path: str, save_path: str):

    # load original

    tensors = {}

    with safe_open(load_path, framework="pt") as f:

        for key in f.keys():

            tensors[key] = f.get_tensor(key)

    # migrate

    new_tensors = {}

    for k, v in tensors.items():

        new_key = k

        # only fix the vae

        if "first_stage_model." in k:

            # migrate q, k, v keys

            new_key = new_key.replace(".to_q.weight", ".q.weight")

            new_key = new_key.replace(".to_q.bias", ".q.bias")

            new_key = new_key.replace(".to_k.weight", ".k.weight")

            new_key = new_key.replace(".to_k.bias", ".k.bias")

            new_key = new_key.replace(".to_v.weight", ".v.weight")

            new_key = new_key.replace(".to_v.bias", ".v.bias")

        new_tensors[new_key] = v

    # save

    save_file(new_tensors, save_path)

if name == "__main__":

    fix_diffusers_model_conversion(

        os.path.normpath(sys.argv[1]),

        os.path.normpath(sys.argv[1] + "_fixed"),

    )
```