AttributeError: 'collections.OrderedDict' object has no attribute 'named_children'

reinforcement-learning
1

import numpy as np

import torch

from data_load import NpyDataset

from main import Generator

generator = Generator(channels=15)

checkpoint = torch.load(‘netd_0.pth’)

generator.load_state_dict(checkpoint, strict=False)

generator.eval()

device = torch.device(‘cpu’)

fix_noises = torch.randn(64, 100, 4, 4).to(device)

dataset = NpyDataset(‘10000_npy/’)

fix_fake_image = generator(fix_noises)

fix_fake_image = (fix_fake_image * (dataset.max - dataset.min)) + dataset.min

fix_fake_image = fix_fake_image[:64].detach().cpu().numpy()

np.save(‘PERM/perm_1’, fix_fake_image)

RuntimeError: Error(s) in loading state_dict for Generator:
size mismatch for main_module.0.weight: copying a param with shape torch.Size([256, 15, 4, 5]) from checkpoint, the shape in current model is torch.Size([100, 1024, 4, 3]).
size mismatch for main_module.0.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([1024]).
size mismatch for main_module.1.weight: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([1024]).
size mismatch for main_module.1.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([1024]).
size mismatch for main_module.1.running_mean: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([1024]).
size mismatch for main_module.1.running_var: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([1024]).
size mismatch for main_module.3.weight: copying a param with shape torch.Size([512, 256, 4, 5]) from checkpoint, the shape in current model is torch.Size([1024, 512, 4, 3]).
size mismatch for main_module.6.weight: copying a param with shape torch.Size([1024, 512, 4, 5]) from checkpoint, the shape in current model is torch.Size([512, 256, 4, 2]).
size mismatch for main_module.6.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([256]).
size mismatch for main_module.7.weight: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([256]).
size mismatch for main_module.7.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([256]).
size mismatch for main_module.7.running_mean: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([256]).
size mismatch for main_module.7.running_var: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([256]).

2

The error message from the title is raised if you try to call .named_children() on a state_dict instead of the model object. The errors mentioned in the post are raised since the model architecture and state_dict parameters do not match in shape.