wandering007 / nasnet-pytorch Goto Github PK

Hi
I seek help as I am unable to reproduce your reported results. Here is my setup:
Python version 3.6.5
Pytorch version 1.1.0

Imagenet validation dataset downloaded from imagenet website and placed in a val folder having the following structure:

imagenet_dataset
    val/
        n01440764/
            ...images...
        n01443537/
            ...images...
        .... 1000 such folders

nasnet_a_large.pth weights downloaded from your dropbox link.

I run your script as follows:
python imagenet_eval.py --nas-type large --resume ../nasnet_a_large.pth --gpus 0 1 --data ../../data/imagenet_dataset --batch-size 4

I get Top1 = 17% and Top5 = 4%
Test set[50000]: Top1: 17.00%, Top5: 4.00%, Average loss: 0.8056

Please help.
Thank you

Hi, I just run the nasnet code but get an error at this line:
RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 8 but got size 7 for tensor number 1 in the list.`
I think x_path_1 is one element longer than x_path_2 along the last two dimensions, as at the previous line you stripped the first element on x_path_2. What do you think?

Thanks a lot for your help.

Best,
Xianzhe

I just use below model file and Use ImageNet (year: 2012)
The ported model files are provided: NASNet-A_Mobile_224
But, I got top 1 acc: 53.97%, top5: 32.50% and This seems overfitting.
Is there something settings should I check when I run these things?

I get this error while loading the weights:

RuntimeError                              Traceback (most recent call last)
<ipython-input-32-d82b0689f3c8> in <module>()
----> 1 model=model.load_state_dict(weights)

/opt/conda/lib/python3.6/site-packages/torch/nn/modules/module.py in load_state_dict(self, state_dict, strict)
    767         if len(error_msgs) > 0:
    768             raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
--> 769                                self.__class__.__name__, "\n\t".join(error_msgs)))
    770 
    771     def _named_members(self, get_members_fn, prefix='', recurse=True):

RuntimeError: Error(s) in loading state_dict for NASNet:
	Missing key(s) in state_dict: "aux_features.2.weight", "aux_features.3.weight", "aux_features.3.bias", "aux_features.3.running_mean", "aux_features.3.running_var", "aux_features.5.weight", "aux_features.6.weight", "aux_features.6.bias", "aux_features.6.running_mean", "aux_features.6.running_var", "aux_linear.weight", "aux_linear.bias". 

I defined a new function
def build_and_load_model(path, nas_type): model = NASNetALarge(1001) if nas_type == 'large' else NASNetAMobile(1001) filename_model = os.path.join(path, 'pytorch', nas_type == 'large' and 'nasnet_a_large.pth' or 'nasnet_a_mobile.pth') state_dict = torch.load(filename_model, map_location='cpu') model.load_state_dict(state_dict) return model
Among them, "tf-models/pytorch/nasnet_a_mobile.pth file is from the pth file you provided

The ported model files are provided: NASNet-A_Mobile_224, NASNet-A_large_331.

Now, I met the following error when call build_and_load_model()
RuntimeError: Error(s) in loading state_dict for NASNet: Missing key(s) in state_dict: "aux_features.2.weight", "aux_features.3.running_var", "aux_features.3.bias", "aux_features.3.weight", "aux_features.3.running_mean", "aux_features.5.weight", "aux_features.6.running_var", "aux_features.6.bias", "aux_features.6.weight", "aux_features.6.running_mean", "aux_linear.bias", "aux_linear.weight".
Is it not match that nasnet_a_mobile.pth provided and the model that NASNetAMobile(1001) function defines?