ML.NET Builder Get Started sample throw excepion when Train time longer than 60 sec

[magnusseptim]

System Information

• Model Builder Version: Latest
• Visual Studio Version : VS 2019 Community 16.2.5

Bug description
Model Builder Get Started, on step :
https://dotnet.microsoft.com/learn/ml-dotnet/get-started-tutorial/train
using wikipedia-detox-250-line-data.tsv, when train time is longer or equal than 60 sec,
output shows :

Exception occured while exploring pipelines:
must be in range [0, 0.5)
Parameter name: L2Regularization
System.ArgumentOutOfRangeException: must be in range [0, 0.5)
Parameter name: L2Regularization
at Microsoft.ML.CLI.CodeGenerator.CodeGenerationHelper.GenerateCode()
at Microsoft.ML.CLI.Program.<>c__DisplayClass1_0.

b__0(NewCommandSettings options)
Please see the log file for more info.
Exiting ...

Last tested algorithm visible in output is :

|9 SgdCalibratedBinary 0,7317 0,8104 0,9053 0,7925 1,4 0 |

Steps to Reproduce

1. Right click on project -> Add -> Machine Learning
2. Follow Get Started from : https://dotnet.microsoft.com/learn/ml-dotnet/get-started-tutorial/intro until "Train your model"
3. In "Time to train (seconds)" textbox set >= 60
4. Wait until output shows error

Expected behavior
Same as you train it shorter, no errors but starting next step

Screenshots

[LittleLittleCloud]

Seems to be the ModelBuilder that still use cli... We can try if it will reproduce on the latest release version

[query-wow]

His also happens to me whenever i try to train my Sentiment Analysis module. 10s and it throws this error. This was after the last update.

[mindaugas-j]

in case you training model from csv file - ensure there is no empty line at the end of the file.

[justinormont]

Duplicate of dotnet/machinelearning#567
Fixed by dotnet/machinelearning#579

Please re-open if you still see this issue. It's possible the sweep ranges were not propagated to the sweep ranges used by AutoML.

Side note: AutoML should be directly accessing the sweep ranges of the ML.NET components. Instead it maintains its own copy of the sweep ranges.

[eyvindwa]

I had this exact issue, and the problem was a blank line at the end of my dataset, like @mindaugas-j pointed out - thanks!

[eyvindwa]

Hmm, well, removing the blank line helped for one dataset, but not for another. Seems this problem is not fixed after all:

Exploring multiple ML algorithms and settings to find you the best model for ML task: multiclass-classification For further learning check: https://aka.ms/mlnet-cli ▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓────────────────────────────────────────────────────────────────────────────────────────────────────Best Accuracy: 79,12%, Best Algorithm: LbfgsLogisticRegressionOva, Last Algorithm: LbfgsMaximumEntropyMulti 00:05:04 Exception occured while exploring pipelines: must be in range [0, 0.5)─────────────────────────────────────────────────────────────────────────────────────────────── Parameter name: L2Regularization Please see the log file for more info. Exiting ...

[adepierre]

I just got the exact same problem. Removing the blank line at the end of the file also seems to fix it for one dataset, but then I tried with another and the problem is back.