Cannot parse target values from big dataset (classification 3+)

[PasaOpasen]

System Information:

• Model Builder Version: 16.0.2003.302.
• Visual Studio Version: Community 16.5.4

Describe the bug

It cannot parse target values from big dataset (classification 3+ from 0 to 10, almost 600 000 samples) after than over 30% of samples in normally parsed (by Convert.ToDouble) dataset.

This error begins only after click at "start training" at 3th block.

Training failed with the exception: System.FormatException: Parsing failed with an exception: Could not parse value 3 in line 186170, column open_channels ---> System.InvalidOperationException: Could not parse value 3 in line 186170, column open_channels в Microsoft.ML.Data.TextLoader.Parser.ProcessOne(FieldSet vs, ColInfo info, ColumnPipe v, Int32 irow, Int64 line) в Microsoft.ML.Data.TextLoader.Parser.ProcessItems(RowSet rows, Int32 irow, Boolean[] active, FieldSet fields, Int32 srcLim, Int64 line) в Microsoft.ML.Data.TextLoader.Parser.ParseRow(RowSet rows, Int32 irow, Helper helper, Boolean[] active, String path, Int64 line, String text) в Microsoft.ML.Data.TextLoader.Cursor.ParallelState.Parse(Int32 tid) в Microsoft.ML.Data.TextLoader.Cursor.ParallelState.ThreadProc(Object obj)

| Trainer MicroAccuracy MacroAccuracy Duration #Iteration | Could not find input column 'Features' Имя параметра: inputSchema Could not find input column 'Features' Имя параметра: inputSchema Parsing failed with an exception: Could not parse value 2 in line 186178, column open_channels

To Reproduce Try to make classificator for open_channels column in this dataset

[JakeRadMSFT]

Thanks for reporting! We'll look into what might be causing this issue.

[beccamc]

@PasaOpasen Can you please download our new build and try this again? I think this is fixed! https://marketplace.visualstudio.com/items?itemName=MLNET.07

[beccamc]

I've tested, and this works on our new build: 16.6.0. Please download and try it out - https://marketplace.visualstudio.com/items?itemName=MLNET.07.

I'm closing this issue. Thank you for reporting!