Open hollowdrutt opened 3 years ago
@LittleLittleCloud Can you take a look at this? Maybe related to the templates?
My best guess is that some column name contains blank space which causes column name doesn't match. @hollowdrutt Can you
@LittleLittleCloud I am using the dataset https://archive.ics.uci.edu/ml/machine-learning-databases/00601/ai4i2020.csv
But I would guess that it is caused by my local culture settings, not by the column names.
Thanks @hollowdrutt We'll take a look
Might relate to #1913
System Information (please complete the following information):
Describe the bug
On which step of the process did you run into an issue: Consume
Clear description of the problem: The generated sample data code snippet does not contain data for all features although the properties are available in the generated ModelInput class. When running the console app the prediction is 1 instead of 0. After adding data for the missing properties the prediction is 0 as expected.
To Reproduce
I have Windows set to use Swedish locale and followed the steps in the learn module Train a predictive maintenance machine learning model with ML.NET Model Builder - Exercise: Train a machine learning model and Exercise: Consume model in .NET console app
Expected behavior I expected the sample code snippet to contain data for all included features and the console application to produce the expected prediction.
Screenshots
Additional context
I have Windows set to use Swedish locale, the csv file with data uses . as decimal separator. The issue is possibly related to Evaluate step can not parse the sample data it has generated, when using Swedish locale