Repository navigation
Parallelism: trouble saturating CPU when making predictions #6702
Description
System Information (please complete the following information):
- OS & Version: Windows Server 2019
- ML.NET Version: ML.NET 1.6.
- .NET Version: .NET 7.0
Describe the bug
I run prediction in Parallel.ForEach loop on 16 core machine. The CPU consumption usually stabilizes near 10-12% even is max parallelism would be 16.
I would like to understand what is might be the bottleneck. Are there some parts of Microsoft.ML that may be limiting, such as locks(?) or parallel data structures that do it internally?
Memory hard Faults usually <10, max < 100. I am caching models and dataset, about 90% RAM used. Disk 0-20% Highest active time.
Activity
dotnet-policy-service commentedon Aug 27, 2026
This issue has been marked
needs-author-actionand may be missing some important information.rosebyte commentedon Aug 27, 2026
Triage: could you please provide more information so we can successfully reproduce the situation?
torronen commentedon Sep 25, 2026
@rosebyte Sorry, it has been a while since I experienced this issue, so I suggest we close this issue. Based on my current knowledge I expect it could have been a GC configuration issue, not ML.NET issue.