Ace-step Lora loader huggingface_hub HFValidationError 59 ACELoRALoader

[b-flip]

When I try to use the Ace-step lora loader with any Ace-step trained lora, also the provided chinese rap lora I got an error from the repo id.

To make sure it wasn't a total path length issue I did the following:

in the ace_step_nodes.py file, I found this:

import folder_paths
cache_dir = folder_paths.get_temp_directory()
models_dir = folder_paths.models_dir
model_path = os.path.join(models_dir, "TTS", "ACE-Step-v1-3.5B")

I changed this to
model_path = os.path.join("C:", "ACE")

and copied the contents of the ACE-Step-v1-3.5B folder to C:\ACE\ and it loads normally in ComfyUI

Still I get the same message, and I think it may be needing a specified repo_id.

Does it have anything to do with this:
https://discuss.huggingface.co/t/hfvalidationerror-repo-id-must-be-in-the-form-repo-name/71171/3

The lora's do load and work in the ComfyUI native lora loader, but of course I would like to use the Ace-Step lora loader to have it work with the Ace-step nodes.

[b-flip]

I think I found it, and it seems the problem was by a path problem within ace_step_nodes.py not matching my structure.

on line 346, this is the code in the file:
models[1].load_lora_adapter(
os.path.join(lora_path, "pytorch_lora_weights.safetensors"),
adapter_name="ace_step_lora",
with_alpha=True,

It seems to expect a folder first (which has the lora name), and then adds the "pytorch_lora_weights.safetensors" hardcoded, so you should not rename the safetensors file. The config.json should also be in the folder.

it doesn't throw an error anymore running in ComfyUI

[b-flip]

It's working now, but it doesn't seem it takes the lora weight into consideration. When I load the lora in the ComfyUI lora loader and use k-sampler with the Ace checkpoint I do instantly hear the lora working.

However no matter what weight I give in the Ace-step lora, the Ace-step node processing doesn't seem to consider it.