EfficientNet-PyTorch
EfficientNet-PyTorch copied to clipboard
lukemelas
model predict error
train is ok
python main.py data/hotDog -b 128 --image_size 300 -a efficientnet-b0 --epochs 300 --gpu 0
predict error
import torch
from torchvision import transforms
from PIL import Image
from efficientnet_pytorch import EfficientNet
tfms = transforms.Compose([transforms.Resize(224), transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), ])
# wget https://www.kaggle.com/yasserhessein/hotdogornothotdog-dataset
img = tfms(
Image.open('data/hotDog/test/7e8ae4cd90a84f1b799e5f6e170bfa4a665dda697ffe3c848f4c644bf224939a.png')).unsqueeze(0)
model = EfficientNet.from_pretrained('efficientnet-b0', weights_path=None, advprop=False, in_channels=3, num_classes=2)
state_dict = torch.load('model_best.pth.tar')
model.load_state_dict(state_dict)
model.cuda()
img = img.cuda()
labels_map = ['hot', 'not_hot']
model.eval()
with torch.no_grad():
outputs = model(img)
for idx in torch.topk(outputs, k=2).indices.squeeze(0).tolist():
prob = torch.softmax(outputs, dim=1)[0, idx].item()
print(dict(label=labels_map[idx], score=prob))
error msg
File "/home/gpu1/.pycharm_helpers/pydev/_pydev_imps/_pydev_execfile.py", line 18, in execfile
exec(compile(contents+"\n", file, 'exec'), glob, loc)
File "/home/kang/coding/demos/EfficientNet-PyTorch-master/examples/imagenet/pre.py", line 28, in <module>
model.load_state_dict(state_dict)
File "/opt/conda/envs/kang-ssd/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1224, in load_state_dict
self.__class__.__name__, "\n\t".join(error_msgs)))
RuntimeError: Error(s) in loading state_dict for EfficientNet:
Missing key(s) in state_dict: "_conv_stem.weight", "_bn0.weight", "_bn0.bias", "_bn0.running_mean", "_bn0.running_var", "_blocks.0._depthwise_conv.weight", "_blocks.0._bn1.weight", "_blocks.0._bn1.bias", "_blocks.0._bn1.running_mean", "_blocks.0._bn1.running_var", "_blocks.0._se_reduce.weight", "_blocks.0._se_reduce.bias", "_blocks.0._se_expand.weight", "_blocks.0._se_expand.bias", "_blocks.0._project_conv.weight", "_blocks.0._bn2.weight", "_blocks.0._bn2.bias", "_blocks.0._bn2.running_mean", "_blocks.0._bn2.running_var", "_blocks.1._expand_conv.weight", "_blocks.1._bn0.weight", "_blocks.1._bn0.bias", "_blocks.1._bn0.running_mean", "_blocks.1._bn0.running_var", "_blocks.1._depthwise_conv.weight", "_blocks.1._bn1.weight", "_blocks.1._bn1.bias", "_blocks.1._bn1.running_mean", "_blocks.1._bn1.running_var", "_blocks.1._se_reduce.weight", "_blocks.1._se_reduce.bias", "_blocks.1._se_expand.weight", "_blocks.1._se_expand.bias", "_blocks.1._project_conv.weight", "_blocks.1._bn2.weight", "_blocks.1._bn2.bias", "_blocks.1._bn2.running_mean", "_blocks.1._bn2.running_var", "_blocks.2._expand_conv.weight", "_blocks.2._bn0.weight", "_blocks.2._bn0.bias", "_blocks.2._bn0.running_mean", "_blocks.2._bn0.running_var", "_blocks.2._depthwise_conv.weight", "_blocks.2._bn1.weight", "_blocks.2._bn1.bias", "_blocks.2._bn1.running_mean", "_blocks.2._bn1.running_var", "_blocks.2._se_reduce.weight", "_blocks.2._se_reduce.bias", "_blocks.2._se_expand.weight", "_blocks.2._se_expand.bias", "_blocks.2._project_conv.weight", "_blocks.2._bn2.weight", "_blocks.2._bn2.bias", "_blocks.2._bn2.running_mean", "_blocks.2._bn2.running_var", "_blocks.3._expand_conv.weight", "_blocks.3._bn0.weight", "_blocks.3._bn0.bias", "_blocks.3._bn0.running_mean", "_blocks.3._bn0.running_var", "_blocks.3._depthwise_conv.weight", "_blocks.3._bn1.weight", "_blocks.3._bn1.bias", "_blocks.3._bn1.running_mean", "_blocks.3._bn1.running_var", "_blocks.3._se_reduce.weight", "_blocks.3._se_reduce.bias", "_blocks.3._se_expand.weight", "_blocks.3._se_expand.bias", "_blocks.3._project_conv.weight", "_blocks.3._bn2.weight", "_blocks.3._bn2.bias", "_blocks.3._bn2.running_mean", "_blocks.3._bn2.running_var", "_blocks.4._expand_conv.weight", "_blocks.4._bn0.weight", "_blocks.4._bn0.bias", "_blocks.4._bn0.running_mean", "_blocks.4._bn0.running_var", "_blocks.4._depthwise_conv.weight", "_blocks.4._bn1.weight", "_blocks.4._bn1.bias", "_blocks.4._bn1.running_mean", "_blocks.4._bn1.running_var", "_blocks.4._se_reduce.weight", "_blocks.4._se_reduce.bias", "_blocks.4._se_expand.weight", "_blocks.4._se_expand.bias", "_blocks.4._project_conv.weight", "_blocks.4._bn2.weight", "_blocks.4._bn2.bias", "_blocks.4._bn2.running_mean", "_blocks.4._bn2.running_var", "_blocks.5._expand_conv.weight", "_blocks.5._bn0.weight", "_blocks.5._bn0.bias", "_blocks.5._bn0.running_mean", "_blocks.5._bn0.running_var", "_blocks.5._depthwise_conv.weight", "_blocks.5._bn1.weight", "_blocks.5._bn1.bias", "_blocks.5._bn1.running_mean", "_blocks.5._bn1.running_var", "_blocks.5._se_reduce.weight", "_blocks.5._se_reduce.bias", "_blocks.5._se_expand.weight", "_blocks.5._se_expand.bias", "_blocks.5._project_conv.weight", "_blocks.5._bn2.weight", "_blocks.5._bn2.bias", "_blocks.5._bn2.running_mean", "_blocks.5._bn2.running_var", "_blocks.6._expand_conv.weight", "_blocks.6._bn0.weight", "_blocks.6._bn0.bias", "_blocks.6._bn0.running_mean", "_blocks.6._bn0.running_var", "_blocks.6._depthwise_conv.weight", "_blocks.6._bn1.weight", "_blocks.6._bn1.bias", "_blocks.6._bn1.running_mean", "_blocks.6._bn1.running_var", "_blocks.6._se_reduce.weight", "_blocks.6._se_reduce.bias", "_blocks.6._se_expand.weight", "_blocks.6._se_expand.bias", "_blocks.6._project_conv.weight", "_blocks.6._bn2.weight", "_blocks.6._bn2.bias", "_blocks.6._bn2.running_mean", "_blocks.6._bn2.running_var", "_blocks.7._expand_conv.weight", "_blocks.7._bn0.weight", "_blocks.7._bn0.bias", "_blocks.7._bn0.running_mean", "_blocks.7._bn0.running_var", "_blocks.7._depthwise_conv.weight", "_blocks.7._bn1.weight", "_blocks.7._bn1.bias", "_blocks.7._bn1.running_mean", "_blocks.7._bn1.running_var", "_blocks.7._se_reduce.weight", "_blocks.7._se_reduce.bias", "_blocks.7._se_expand.weight", "_blocks.7._se_expand.bias", "_blocks.7._project_conv.weight", "_blocks.7._bn2.weight", "_blocks.7._bn2.bias", "_blocks.7._bn2.running_mean", "_blocks.7._bn2.running_var", "_blocks.8._expand_conv.weight", "_blocks.8._bn0.weight", "_blocks.8._bn0.bias", "_blocks.8._bn0.running_mean", "_blocks.8._bn0.running_var", "_blocks.8._depthwise_conv.weight", "_blocks.8._bn1.weight", "_blocks.8._bn1.bias", "_blocks.8._bn1.running_mean", "_blocks.8._bn1.running_var", "_blocks.8._se_reduce.weight", "_blocks.8._se_reduce.bias", "_blocks.8._se_expand.weight", "_blocks.8._se_expand.bias", "_blocks.8._project_conv.weight", "_blocks.8._bn2.weight", "_blocks.8._bn2.bias", "_blocks.8._bn2.running_mean", "_blocks.8._bn2.running_var", "_blocks.9._expand_conv.weight", "_blocks.9._bn0.weight", "_blocks.9._bn0.bias", "_blocks.9._bn0.running_mean", "_blocks.9._bn0.running_var", "_blocks.9._depthwise_conv.weight", "_blocks.9._bn1.weight", "_blocks.9._bn1.bias", "_blocks.9._bn1.running_mean", "_blocks.9._bn1.running_var", "_blocks.9._se_reduce.weight", "_blocks.9._se_reduce.bias", "_blocks.9._se_expand.weight", "_blocks.9._se_expand.bias", "_blocks.9._project_conv.weight", "_blocks.9._bn2.weight", "_blocks.9._bn2.bias", "_blocks.9._bn2.running_mean", "_blocks.9._bn2.running_var", "_blocks.10._expand_conv.weight", "_blocks.10._bn0.weight", "_blocks.10._bn0.bias", "_blocks.10._bn0.running_mean", "_blocks.10._bn0.running_var", "_blocks.10._depthwise_conv.weight", "_blocks.10._bn1.weight", "_blocks.10._bn1.bias", "_blocks.10._bn1.running_mean", "_blocks.10._bn1.running_var", "_blocks.10._se_reduce.weight", "_blocks.10._se_reduce.bias", "_blocks.10._se_expand.weight", "_blocks.10._se_expand.bias", "_blocks.10._project_conv.weight", "_blocks.10._bn2.weight", "_blocks.10._bn2.bias", "_blocks.10._bn2.running_mean", "_blocks.10._bn2.running_var", "_blocks.11._expand_conv.weight", "_blocks.11._bn0.weight", "_blocks.11._bn0.bias", "_blocks.11._bn0.running_mean", "_blocks.11._bn0.running_var", "_blocks.11._depthwise_conv.weight", "_blocks.11._bn1.weight", "_blocks.11._bn1.bias", "_blocks.11._bn1.running_mean", "_blocks.11._bn1.running_var", "_blocks.11._se_reduce.weight", "_blocks.11._se_reduce.bias", "_blocks.11._se_expand.weight", "_blocks.11._se_expand.bias", "_blocks.11._project_conv.weight", "_blocks.11._bn2.weight", "_blocks.11._bn2.bias", "_blocks.11._bn2.running_mean", "_blocks.11._bn2.running_var", "_blocks.12._expand_conv.weight", "_blocks.12._bn0.weight", "_blocks.12._bn0.bias", "_blocks.12._bn0.running_mean", "_blocks.12._bn0.running_var", "_blocks.12._depthwise_conv.weight", "_blocks.12._bn1.weight", "_blocks.12._bn1.bias", "_blocks.12._bn1.running_mean", "_blocks.12._bn1.running_var", "_blocks.12._se_reduce.weight", "_blocks.12._se_reduce.bias", "_blocks.12._se_expand.weight", "_blocks.12._se_expand.bias", "_blocks.12._project_conv.weight", "_blocks.12._bn2.weight", "_blocks.12._bn2.bias", "_blocks.12._bn2.running_mean", "_blocks.12._bn2.running_var", "_blocks.13._expand_conv.weight", "_blocks.13._bn0.weight", "_blocks.13._bn0.bias", "_blocks.13._bn0.running_mean", "_blocks.13._bn0.running_var", "_blocks.13._depthwise_conv.weight", "_blocks.13._bn1.weight", "_blocks.13._bn1.bias", "_blocks.13._bn1.running_mean", "_blocks.13._bn1.running_var", "_blocks.13._se_reduce.weight", "_blocks.13._se_reduce.bias", "_blocks.13._se_expand.weight", "_blocks.13._se_expand.bias", "_blocks.13._project_conv.weight", "_blocks.13._bn2.weight", "_blocks.13._bn2.bias", "_blocks.13._bn2.running_mean", "_blocks.13._bn2.running_var", "_blocks.14._expand_conv.weight", "_blocks.14._bn0.weight", "_blocks.14._bn0.bias", "_blocks.14._bn0.running_mean", "_blocks.14._bn0.running_var", "_blocks.14._depthwise_conv.weight", "_blocks.14._bn1.weight", "_blocks.14._bn1.bias", "_blocks.14._bn1.running_mean", "_blocks.14._bn1.running_var", "_blocks.14._se_reduce.weight", "_blocks.14._se_reduce.bias", "_blocks.14._se_expand.weight", "_blocks.14._se_expand.bias", "_blocks.14._project_conv.weight", "_blocks.14._bn2.weight", "_blocks.14._bn2.bias", "_blocks.14._bn2.running_mean", "_blocks.14._bn2.running_var", "_blocks.15._expand_conv.weight", "_blocks.15._bn0.weight", "_blocks.15._bn0.bias", "_blocks.15._bn0.running_mean", "_blocks.15._bn0.running_var", "_blocks.15._depthwise_conv.weight", "_blocks.15._bn1.weight", "_blocks.15._bn1.bias", "_blocks.15._bn1.running_mean", "_blocks.15._bn1.running_var", "_blocks.15._se_reduce.weight", "_blocks.15._se_reduce.bias", "_blocks.15._se_expand.weight", "_blocks.15._se_expand.bias", "_blocks.15._project_conv.weight", "_blocks.15._bn2.weight", "_blocks.15._bn2.bias", "_blocks.15._bn2.running_mean", "_blocks.15._bn2.running_var", "_conv_head.weight", "_bn1.weight", "_bn1.bias", "_bn1.running_mean", "_bn1.running_var", "_fc.weight", "_fc.bias".
Unexpected key(s) in state_dict: "epoch", "arch", "state_dict", "best_acc1", "optimizer".
terminate called without an active exception
predict error
import torch
from torchvision import transforms
from PIL import Image
from efficientnet_pytorch import EfficientNet
tfms = transforms.Compose([transforms.Resize(224), transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), ])
# wget https://www.kaggle.com/yasserhessein/hotdogornothotdog-dataset
img = tfms(
Image.open('data/hotDog/test/7e8ae4cd90a84f1b799e5f6e170bfa4a665dda697ffe3c848f4c644bf224939a.png')).unsqueeze(0)
# this change weights_path=
model = EfficientNet.from_pretrained('efficientnet-b0', weights_path='model_best.pth.tar', advprop=False, in_channels=3, num_classes=2)
# state_dict = torch.load('model_best.pth.tar')
# model.load_state_dict(state_dict)
model.cuda()
img = img.cuda()
labels_map = ['hot', 'not_hot']
model.eval()
with torch.no_grad():
outputs = model(img)
for idx in torch.topk(outputs, k=2).indices.squeeze(0).tolist():
prob = torch.softmax(outputs, dim=1)[0, idx].item()
print(dict(label=labels_map[idx], score=prob))
error
model = EfficientNet.from_pretrained('efficientnet-b0', weights_path='model_best.pth.tar', advprop=False, in_channels=3, num_classes=2)
File "/home/kang/coding/demos/EfficientNet-PyTorch-master/efficientnet_pytorch/model.py", line 379, in from_pretrained
load_fc=(num_classes == 1000), advprop=advprop)
File "/home/kang/coding/demos/EfficientNet-PyTorch-master/efficientnet_pytorch/utils.py", line 608, in load_pretrained_weights
state_dict.pop('_fc.weight')
KeyError: '_fc.weight'
