Repository navigation
Text classification : Target 4294967295 is out of bounds #6534
Description
System Information (please complete the following information):
- ML.NET Version: 2.0
Describe the bug
Detailed info can be found in
#dotnet/machinelearning-modelbuilder#2369
Activity
LittleLittleCloud commentedon Jan 5, 2023
It seems that if label's value is
NaNor any values that not appear in ValueToKey term mapper, that label will be mapped to0.machinelearning/src/Microsoft.ML.Data/Transforms/ValueToKeyMappingTransformerImpl.cs
Line 720 in 9d798f1
So maybe a fix can be proccessing data and filter out the rows where label is null before training in text classification?
michaelgsharp commentedon Feb 6, 2023
Yeah, the torch sharp model is 1 based. I don't know if they have an "unknown" value or not. If they do we need to map 0 to that correctly. If not we need to filter it out before we pass it.
yurant333 commentedon Jun 21, 2024
Hi guys!
Please clarify how you can temporarily solve the problem?
To work with labels, I made an enum in advance, where the elements began with both 0 and 1.
When starting from 0 I get the error:
System.Runtime.InteropServices.ExternalException (0x80004005): Target 4294967295 is out of bounds.
When starting from 1 I get the error:
System.Runtime
System.Runtime.InteropServices.ExternalException (0x80004005): Target 39 is out of bounds.
LittleLittleCloud commentedon Jun 21, 2024
The fix on our side is in dataset filter out the labels where value is empty or null