diff options
author | Lester Solbakken <lester.solbakken@gmail.com> | 2024-05-30 15:14:18 +0200 |
---|---|---|
committer | Lester Solbakken <lester.solbakken@gmail.com> | 2024-05-30 15:14:18 +0200 |
commit | f26240553b4cfddacfff667c87a0768c450b0908 (patch) | |
tree | 1ff64311738154ef9bae02177591645eb347ac3e | |
parent | 168311c0e9832c31f371bc56d52c6d09dca89378 (diff) |
Use chat template in llama if available
-rw-r--r-- | model-integration/src/main/java/ai/vespa/llm/clients/LocalLLM.java | 2 |
1 files changed, 2 insertions, 0 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 fd02756e2ea..bbb82db7139 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 @@ -114,6 +114,8 @@ public class LocalLLM extends AbstractComponent implements LanguageModel { options.ifPresent("repeatpenalty", (v) -> inferParams.setRepeatPenalty(Float.parseFloat(v))); // Todo: more options? + inferParams.setUseChatTemplate(true); + var completionFuture = new CompletableFuture<Completion.FinishReason>(); var hasStarted = new AtomicBoolean(false); try { |