HTTP Error 503: Service Unavailable

[sahilpalarpwar]

I am facing this error while downloading MNIST in google colab.

[birajde]

Http error 503 means that the website you are trying to download from is down.

[yogeshbadhe257 (Yogesh)]

Yes, but without downloading that dataset, the tutorial cannot be completed. Is there a work around?
If Jovian people have those files, they need to preload it in colab.

[birajde]

Check this forum post given below

Lecture 2: Working with Images & Logistic Regression

@ccxuan123 @2019uec1805 You might have to do a manual download: It seems the website you are trying to download it from is down.

If you are still unable to download wait for it, I will help you with the correct code soon. Sorry for the inconvenience.

[birajde]

@yogeshbadhe257 Were you able to resolve your issue? If you haven’t check this forum post below

Update - Lecture 2 Notebook of Deep Learning with Pytorch course

How to avoid HTTP-Error 503 while downloading the MNIST Dataset If you are facing HTTP-Error 503 while downloading the MNIST Dataset it may be due to the website which is temporarily or permanently unavailable. To avoid this error you have to update the following code from the notebook. Code to be Updated(where you are getting the HTTP Error): # Download training dataset dataset = MNIST(root='data/', download=True) Updated code !wget www.di.ens.fr/~lelarge/MNIST.tar.gz !tar -zxvf MNIST.tar.gz …

[cmeher123 (Chinmaya Kumar Meher)]

I have downloaded the MNIST dataset earlier while running the notebook in kaggle. After someday when I try to run the code again. The dataset is downloading again. Do we need to download again and again.

[Sebgolos (Sebastian)]

The notebooks on kaggle follow the same principle as those run on binder - they’re “turned off” when you don’t work on them (albeit kaggle allows for a bit more inactivity I think).

When you go back to the notebook after some time, the old data is gone (to free up the resources of unused notebooks), so it has to be downloaded again.

Solutions:

• work locally if you don’t want to download the data again
• kaggle has an option to add a dataset to a notebook, so it doesn’t have to be downloaded. There are many datasets, and I’m 98% sure that there’s MNIST as well.