Error when running main.py about maml HOT 6 CLOSED

Dear,

How to start the project ?

I just want to test the regression using MAML in predicting the price of stock.

Any help?

from maml.

It seems like it is not finding the images, causing folders to be an empty list. It looks for the images in ./data/omniglot_resized (on this line of code https://github.com/cbfinn/maml/blob/master/data_generator.py#L40)

Are you calling the code from the maml/ directory? e.g. python main.py
Can you see if the folders is indeed an empty list?

from maml.

Yes, I was running main.py from maml/ directory,
I found that when I run main.py, it was able to execute the line https://github.com/cbfinn/maml/blob/master/main.py#L272, and generate the tensor shape (32, 10, 784) (32, 10, 784) for image_tensor, label_tensor respectively , and the folders in https://github.com/cbfinn/maml/blob/master/data_generator.py#L95 was not a empty list.

but when it continue and execute the line https://github.com/cbfinn/maml/blob/master/main.py#L280, it print out the error message I describe above, and the folders in https://github.com/cbfinn/maml/blob/master/data_generator.py#L95 was empty list when I print len(folders), it shows zero.

So it was able to generate image_tensor when calling data_generator.make_data_tensor(), but not data_generator.make_data_tensor(train=False)

Thank you for your response, really looking forward to test and play with the model !
Max.

from maml.

Ah, okay. Make sure you download both images_background and images_evaluation from the omniglot website, and put them in the same directory (and run the preprocessing for all images)

That should solve your problem. (It seems to be that there are <= 1200 character classes in the dataset that you downloaded)

If this solves your issue, then I will clarity the Omniglot data instructions.

from maml.

Yes! This solves the problem. Thank you very much.

from maml.

Hi, I am having a similar problem. I can't find any dataset to download on the omniglot website, so I got them from here: https://github.com/brendenlake/omniglot

Where can we get the right dataset?

from maml.