opparco / stable-diffusion-webui-composable-lora Goto Github PK

I am running a Jan commit of Auto1111, so maybe its the cause but I do get the following error:
File "/sd/modules/extra_networks.py", line 75, in activate
extra_network.activate(p, extra_network_args)
File "/sd/extensions-builtin/Lora/extra_networks_lora.py", line 17, in activate
lora.load_loras(names, multipliers)
File "/sd/extensions-builtin/Lora/lora.py", line 146, in load_loras
lora = load_lora(name, lora_on_disk.filename)
File "/sd/extensions-builtin/Lora/lora.py", line 98, in load_lora
module = torch.nn.Conv2d(weight.shape[1], weight.shape[0], (1, 1), bias=False)
IndexError: tuple index out of range

When I duplicate lora, the image becomes corrupted, both with and without two-shot.


Without two-shot

With two-shot

@opparco
AUTOMATIC1111/stable-diffusion-webui-extensions#79
we have received from reports that this extension is in a broken state no longer maintained
and based on the current open issues this is the case
since apparently a working Fork is available from @a2569875
we intend to redirect the git url on webui extensions index to direct to @a2569875 fork if the issues it's not resolved in 1~2 week

@a2569875 are you ok with us redirecting to your repo? or you prefer us not to?

hi,
I just tested this extension, and it works with lora, but when I tried with lycoris, it always caused this error:

...
File "D:\AI Arts\Automatic1111 Webui\stable-diffusion-webui\extensions\stable-diffusion-webui-composable-lora\composable_lycoris.py", line 181, in inference
    return self.op(x, self.weight, **self.extra_args)
RuntimeError: Input type (torch.cuda.HalfTensor) and weight type (torch.HalfTensor) should be the same

anyone else has this problem?
thanks

Causing all types of issues.

that is a bug in extension stable-diffusion-webui-composable-lora as it hasn't been updated in a while
it tries to read option lora_apply_to_outputs, but that option was removed from core repo last month:
AUTOMATIC1111/stable-diffusion-webui@80b26d2

and in general, extension should never assume that any user option exists, it should read it as optional value.

vladmandic/automatic#225

It would appear that when this extension is enabled, I am unable to run generations where the batch size is greater than 1, giving an error like IndexError: list index out of range.

Is this an intentional/known limitation?

Make it works that way but with BREAK keyword, because it much more useful than AND

Hello. Just tried this extension, and it's awesome. Tried with built-in Lora, and it worked very well.

Here's a retro pin up Lora I trained on SD 2.1, generated on Illuminati Diffusion:
Composable Lora disabled:

Composable Lora enabled:

Some of my Loras do not work using the built-in Lora, so I use this extension by kohya:

https://github.com/kohya-ss/sd-webui-additional-networks

Using this, Composable Lora does not seem to work at all.

Composable Lora enabled:

Hi,

Seeing this error in newer commits here
Any way to bypass it? I saw other repos patched it, but perhaps Vlad has a point and it should be looked at here, as the check is problematic

I'm using automatic1111 API to call txt2img/img2img process and would like to take the advantage of composable lora extension. But I have no idea how to enable the extension through API.

Need some help,
Thanks a lot!

How to start the red lora icon? I do not have this icon when I have installed it, but my classmate has this icon when he has installed it. Could you please help me

It just don't make a change, the images are same on or off.

Could you provide an example of the LORA applying to only a subprompt?

I'm attempting to use this along with the two shot diffusion port but the LORAs seem to still be applying to the whole prompt.