diffusers 出现TypeError: For single controlnet: `controlnet_conditioning_scale` must be type `float`.

测试ComfyUI-InstantID 时出现如下

[2024-04-12 00:37] !!! Exception during processing !!!
[2024-04-12 00:37] Traceback (most recent call last):
  File "/mnt/data/share/aigc/ComfyUI/execution.py", line 177, in recursive_execute
    output_data, output_ui = get_output_data(obj, input_data_all)
  File "/mnt/data/share/aigc/ComfyUI/execution.py", line 81, in get_output_data
    return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
  File "/mnt/data/share/aigc/ComfyUI/execution.py", line 74, in map_node_over_list
    results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
  File "/mnt/data/share/aigc/ComfyUI/custom_nodes/ComfyUI-InstantID/InstantIDNode.py", line 287, in id_generate_image
    output = pipe(
  File "/mnt/data/share/aigc/ComfyUI/venv/Comfyui/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
    return func(*args, **kwargs)
  File "/mnt/data/share/aigc/ComfyUI/custom_nodes/ComfyUI-InstantID/pipeline_stable_diffusion_xl_instantid.py", line 438, in __call__
    self.check_inputs(
  File "/mnt/data/share/aigc/ComfyUI/venv/Comfyui/lib/python3.10/site-packages/diffusers/pipelines/controlnet/pipeline_controlnet_sd_xl.py", line 685, in check_inputs
    TypeError: For single controlnet: `controlnet_conditioning_scale` must be type `float`.

经过定位，应该是diffusers的版本差异引发的，问题为参数传递问题：

pipeline_controlnet_sd_xl.py文件中的定义为：

    def check_inputs(
        self,
        prompt,
        prompt_2,
        image,
        callback_steps,
        negative_prompt=None,
        negative_prompt_2=None,
        prompt_embeds=None,
        negative_prompt_embeds=None,
        pooled_prompt_embeds=None,
        ip_adapter_image=None,
        ip_adapter_image_embeds=None,
        negative_pooled_prompt_embeds=None,
        controlnet_conditioning_scale=1.0,
        control_guidance_start=0.0,
        control_guidance_end=1.0,
        callback_on_step_end_tensor_inputs=None,
    ):
 

而pipeline_stable_diffusion_xl_instantid.py文件中的调用如下 （大约在453行）

        self.check_inputs(
            prompt,
            prompt_2,
            image,
            callback_steps,
            negative_prompt,
            negative_prompt_2,
            prompt_embeds,
            negative_prompt_embeds,
            pooled_prompt_embeds,
            negative_pooled_prompt_embeds,
            controlnet_conditioning_scale,
            control_guidance_start,
            control_guidance_end,
            callback_on_step_end_tensor_inputs,
        )

bug修复方式为：将pipeline_stable_diffusion_xl_instantid.py调用部分修改为：

        self.check_inputs(
            prompt,
            prompt_2,
            image,
            callback_steps,
            negative_prompt,
            negative_prompt_2,
            prompt_embeds,
            negative_prompt_embeds,
            pooled_prompt_embeds,
            negative_pooled_prompt_embeds,
            controlnet_conditioning_scale = controlnet_conditioning_scale,
            control_guidance_start = control_guidance_start,
            control_guidance_end = control_guidance_end,
            callback_on_step_end_tensor_inputs = callback_on_step_end_tensor_inputs,
        )

问题解决
————————————————

                            版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
                        
原文链接：https://blog.csdn.net/chawk1978/article/details/137665577