Merge pull request #30988 from vespa-engine/lesters/llm-minimum-threads

Set minimum number of threads to 1

1 files changed, 1 insertions, 2 deletions

diff --git a/model-integration/src/main/java/ai/vespa/llm/clients/LocalLLM.java b/model-integration/src/main/java/ai/vespa/llm/clients/LocalLLM.java
index fd1b8b700c8..aa7c071b93a 100644
--- a/model-integration/src/main/java/ai/vespa/llm/clients/LocalLLM.java
+++ b/model-integration/src/main/java/ai/vespa/llm/clients/LocalLLM.java
@@ -9,7 +9,6 @@ import com.yahoo.component.AbstractComponent;
 import com.yahoo.component.annotation.Inject;
 import de.kherud.llama.LlamaModel;
 import de.kherud.llama.ModelParameters;
-import de.kherud.llama.args.LogFormat;
 
 import java.util.ArrayList;
 import java.util.List;
@@ -43,7 +42,7 @@ public class LocalLLM extends AbstractComponent implements LanguageModel {
         maxTokens = config.maxTokens();
 
         // Only used if GPU is not used
-        var defaultThreadCount = Runtime.getRuntime().availableProcessors() - 2;
+        var defaultThreadCount = Math.max(Runtime.getRuntime().availableProcessors() - 2, 1);
 
         var modelFile = config.model().toFile().getAbsolutePath();
         var modelParams = new ModelParameters()