Running into ValueError: NaTType does not support utcoffset For 2m_random_50k dataset

#7
by elisachen - opened

Hi!

I'm using colab to read in a random 50k sample from the 2M dataset. However, for image num 46996, the load_dataset function errors during train split due to the following ValueError:

image.png

Please advise on how I can go around this issue.

Thanks!

That's exactly the same error I've encountered just a few minutes ago but for a different split.

Here is the dataset I use:

ds = datasets.load_dataset("poloclub/diffusiondb", "2m_all", split="train")
Generating train split:
1577993/0 [27:03<00:00, 72.85 examples/s]
ValueError: NaTType does not support utcoffset

Would be glad to know about possible fixes. The datasets version is '2.10.1'.

@devforfu Did you manage to resolve this issue?

@maloyan Unfortunately, no. I decided to unpack the downloaded zip archives and use them directly, without datasets interface.

I encountered the exact same issue for random 50k as well as 100k, regardless of whatever machine or environment I have attempted...

If you used the β€œ2m_random_1K” and the try to use another subset (like "2m_random_5k" or "2m_first_1k")under the same cache path (the default path is ".\cache"), the error will happen.
I try to delete all the files of the dataset in my sever, and reload the subset "2m_first_10K". It is ok.

Polo Club of Data Science org
β€’
edited May 9, 2023

Hi all, sorry for the late reply. This should be fixed now with 4de7acd1878f82cc474f4bc432e27f4b3e363ad7. You can try to call load_dataset again. Let me know if it does not fix the issue. Thanks!

xiaohk changed discussion status to closed

Sign up or log in to comment