VanillaCFG.__init__() got an unexpected keyword argument 'dyn_thresh_config'

[YingkaiD]

Traceback (most recent call last): File "/data/yingkai/sdxl/generative-models/tests/inference/test_text_to_image.py", line 23, in i2t = pipeline.text_to_image( File "/home/yingkai/anaconda3/envs/sdxl/lib/python3.10/site-packages/sgm/inference/api.py", line 192, in text_to_image sampler = get_sampler_config(params) File "/home/yingkai/anaconda3/envs/sdxl/lib/python3.10/site-packages/sgm/inference/api.py", line 370, in get_sampler_config return DPMPP2MSampler( File "/home/yingkai/anaconda3/envs/sdxl/lib/python3.10/site-packages/sgm/modules/diffusionmodules/sampling.py", line 32, in init self.guider = instantiate_from_config( File "/home/yingkai/anaconda3/envs/sdxl/lib/python3.10/site-packages/sgm/util.py", line 175, in instantiate_from_config return get_obj_from_str(config["target"])(**config.get("params", dict())) TypeError: VanillaCFG.init() got an unexpected keyword argument 'dyn_thresh_config'

[SWHL]

I met the same error. When I read the source code, I found that the dyn_thresh_config only appears in one place:

https://github.com/Stability-AI/generative-models/blob/fbdc58cab9f4ee2be7a5e1f2e2787ecd9311942f/sgm/inference/api.py#L290-L300

But in the directory of sgm.modules.diffusionmodules.sampling_utils.NoDynamicThresholding, I couldn't find the implementation of this function.

Finally I commented out this parameter as follows, and the program can run successfully.

        guider_config = {
            "target": "sgm.modules.diffusionmodules.guiders.VanillaCFG",
            "params": {"scale": scale},
            # "params": {"scale": scale, "dyn_thresh_config": dyn_thresh_config},
        }