I have been running several independent multi-class NL models on an identical data set (to compare performance to a multi-label model) and had no problems importing the data or running the models. I've just been through the identical preparation process, uploaded the file to the bucket and now get this error on import:
Uri is not found in CSV row ",NotWarm".
Warm and NotWarm are my labels. A sample of the csv is below so you can see the format:
"To ensure you get the best possible service, we stagger the cut-off time for next day delivery from 5pm right up until Midnight.",Warm
You’ll be able to see if Next Day Delivery is still available when you place your order.,NotWarm
"You can choose a home delivery option, which lets you have your order delivered to an address of your choice.",Warm
"Some eligible items also let you choose Click + Collect, where your order is delivered to a local store.",NotWarm
I've double checked all the advice about preparing datasets on the AutoML help pages. The file itself has been encoded in UTF-8 using Notepad++ so there should be nothing amiss with the CSV format. The file is identical to those I've used previously except for the labels.
Has something changed on the AutoML NL process as it was over a month since my last model was created?
Thanks in advance for any guidance.