AutoTrain - unable to upload the dataset

I wanted to try AutoTrain but I’m unable to upload the dataset (a Token Classification task).

After I choose the file to upload there is a modal to choose the Split type (Training / Validation / Auto) and map columns. After clicking “Add to project” button “Uploading your file. This may take up to a few minutes, depending on your connection speed.” message appears and it never disappears. From the Network tab in the browser’s dev tools I can see that there is some kind of CORS error.

AutoTrain project id: AutoTrain - HuggingFace

hello @inouts, indeed your data was not upload. Could you share more details? which browser were you using? what’s the file format and size? thanks

@radames it’s a 193MB CSV file. At first I tried Firefox a few times and then Chrome.

After first (failed) try I’ve noticed that a new dataset was created - but it wasn’t connected to the AutoTrain’s project. Project’s AutoTrain data page was still empty. So I deleted it in the hope that the whole process will be successful in the next try.

There is currently another error when trying to upload a file:

Hello @inouts

When creating a project in AutoTrain, an associated dataset repo is created on the HuggingFace Hub to store your data files.
When you upload a file through AutoTrain, it tries to push it to that dataset repo. Since you deleted it, that dataset repo cannot be found (hence the 404 - not found error).
Creating a dataset repo named ‘autotrain-data-inouts-token-classification’ on the HuggingFace Hub should solve your last issue.

Let me know if it helps!

Thanks @sbrandeis ! I recreated the dataset manually and there is no more 404 error.

We’re back to the first error :slight_smile: :

I plan to convert the dataset to the json format and see if it helps.

So I converted dataset into JSONL format.

Preloaded data shown in the initial modal looked promising (lists detected):

but unfortunately the original error is still there:

(sorry for the second message in a row but I was unable to upload the second screenshot)