What's the input form of conditioning_multi_spectral_image when use controlnet_sd21_md7norm_fmow_condres256

[qqqtwh]

unet_path = '/path/checkpoints/finetune_sd21_sn-satlas-fmow_snr5_md7norm_bs64/checkpoint-100000/'
path = '/path/checkpoints/controlnet_sd21_md7norm_fmow_condres256'

unet = SatUNet.from_pretrained(unet_path, subfolder="unet", num_metadata=7, torch_dtype=torch.float16)
controlnet = ControlNetModel.from_pretrained(path, subfolder="controlnet", num_metadata=7, torch_dtype=torch.float16)
pipe = DiffusionSatControlNetPipeline.from_pretrained(path, unet=unet, controlnet=controlnet, torch_dtype=torch.float16)
pipe = pipe.to("cuda")
....
image = pipe(
caption.replace('naip', 'satlas'), [conditioning_multi_spectral_image],
metadata=metadata, cond_metadata=cond_metadata,
num_inference_steps=50, guidance_scale=1.0, is_temporal=True).images[0]