What is the required file format for Google AutoML Datasets?

Question

Whenever I try to upload my dataset to the AutoML Natural Language Web UI, I get the error

Something is wrong, please try again.

The documentation is not very insightful about how my CSV file is supposed to look, but I tried to make a simple sample file just to make sure it works at all, it looks like this:

text,label
asdf,cat
asodlkao,dog
asdkasdsadksafask,cat
waewq23,cat
dads,cat
saiodjas,cat
skdoaskdoas,dog
hgfkgizk,dog
fzdrgbfd,cat
otiujrhzgf,cat
vchztzr,dog
aksodkasodks,dog
sderftz,dog
dsoakd,dog
qweqweqw,cat
asdqweqe,cat
dkawosdkaodk,dog
ewqeweq,cat
fdsffds,dog
bvcghh,cat
rthnghtd,dog
sdkosadkasodk,cat
sdjidghdfig,cat
kfodskdsof,dog
saodsadok,dog
ksaodksaod,dog
vncvb,cat

I chose this formatting according to the Google suggested Syntax

But even with this formatting I still get the same error I've seen this question Format of the input dataset for Google AutoML Natural Language multi-label text classification but according to the answers there it seems my formatting should work, so I do not know why I get the error

Answer 1

I have found the problem! As Michal K said, there was nothing wrong with the formatting, the real problem was I was not assigned the role of Storage Object Creator, which is necessary because the Data is uploaded in Cloud Storage first

Answer 2

I've just copied the CSV file and uploaded it to my own project and the dataset created worked. One problem is that an extra label was created "label" - this is because the header is not expected to be in the csv file (probably this should get fixed).

Based on that it seems the problem isn't the CSV file format. I would recommend to check if your project is setup correctly. You can open a bug to get someones help. Either you can open a bug in public issue tracker or send feedback using the UI (there is 'Feedback' option in the menu on top right side of the page).