Question about missing key of model weights

[NKUShaw]

Try to run the code: "
import argparse
import torch
import torchvision

from ddpm import script_utils

python ddpmsample.py --model_path='./ddpm_logs/cifar10-ddpm-2024-09-09-02-56-iteration-13000-optim.pth' --save_dir='outputs' --use_labels=True

def main():
args = create_argparser().parse_args()
device = args.device

try:
    diffusion = script_utils.get_diffusion_from_args(args).to(device)
    diffusion.load_state_dict(torch.load(args.model_path))
    if args.use_labels:
        for label in range(10):
            y = torch.ones(args.num_images // 10, dtype=torch.long, device=device) * label
            samples = diffusion.sample(args.num_images // 10, device, y=y)
            
            for image_id in range(len(samples)):
                image = ((samples[image_id] + 1) / 2).clip(0, 1)
                image = torchvision.transforms.Resize((128, 128))(image)  
                torchvision.utils.save_image(image, f"{args.save_dir}/{label}-{image_id}.png")
    else:
        samples = diffusion.sample(args.num_images, device)
        
        for image_id in range(len(samples)):
            image = ((samples[image_id] + 1) / 2).clip(0, 1)
            image = torchvision.transforms.Resize((128, 128))(image) 
            torchvision.utils.save_image(image, f"{args.save_dir}/{image_id}.png")
except KeyboardInterrupt:
    print("Keyboard interrupt, generation finished early")

def create_argparser():
device = torch.device("cuda:2") if torch.cuda.is_available() else torch.device("cpu")
print(device)
defaults = dict(num_images=100,
device=device,
schedule_low=1e-4,
schedule_high=0.02,
)
defaults.update(script_utils.diffusion_defaults())

parser = argparse.ArgumentParser()
parser.add_argument("--model_path", type=str, default='./ddpm_logs/cifar10-ddpm-2024-09-09-02-56-iteration-279000-model.pth')
parser.add_argument("--save_dir", type=str, default='./outputs')
script_utils.add_dict_to_argparser(parser, defaults)
return parser

if name == "main":
main()
"

Traceback (most recent call last):
File "XXX/MyMethod/ddpmsample.py", line 54, in
main()
File "XXX/MyMethod/ddpmsample.py", line 14, in main
diffusion.load_state_dict(torch.load(args.model_path))
File "XXX/lib/python3.9/site-packages/torch/nn/modules/module.py", line 2215, in load_state_dict
raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for GaussianDiffusion:
Missing key(s) in state_dict: "model.downs.0.class_bias.weight", "model.downs.1.class_bias.weight", "model.downs.3.class_bias.weight", "model.downs.4.class_bias.weight", "model.downs.6.class_bias.weight", "model.downs.7.class_bias.weight", "model.downs.9.class_bias.weight", "model.downs.10.class_bias.weight", "model.ups.0.class_bias.weight", "model.ups.1.class_bias.weight", "model.ups.2.class_bias.weight", "model.ups.4.class_bias.weight", "model.ups.5.class_bias.weight", "model.ups.6.class_bias.weight", "model.ups.8.class_bias.weight", "model.ups.9.class_bias.weight", "model.ups.10.class_bias.weight", "model.ups.12.class_bias.weight", "model.ups.13.class_bias.weight", "model.ups.14.class_bias.weight", "model.mid.0.class_bias.weight", "model.mid.1.class_bias.weight", "ema_model.downs.0.class_bias.weight", "ema_model.downs.1.class_bias.weight", "ema_model.downs.3.class_bias.weight", "ema_model.downs.4.class_bias.weight", "ema_model.downs.6.class_bias.weight", "ema_model.downs.7.class_bias.weight", "ema_model.downs.9.class_bias.weight", "ema_model.downs.10.class_bias.weight", "ema_model.ups.0.class_bias.weight", "ema_model.ups.1.class_bias.weight", "ema_model.ups.2.class_bias.weight", "ema_model.ups.4.class_bias.weight", "ema_model.ups.5.class_bias.weight", "ema_model.ups.6.class_bias.weight", "ema_model.ups.8.class_bias.weight", "ema_model.ups.9.class_bias.weight", "ema_model.ups.10.class_bias.weight", "ema_model.ups.12.class_bias.weight", "ema_model.ups.13.class_bias.weight", "ema_model.ups.14.class_bias.weight", "ema_model.mid.0.class_bias.weight", "ema_model.mid.1.class_bias.weight".

what happend