diff --git a/config/config.yaml b/config/config.yaml index 82cb8f2..09a6d19 100644 --- a/config/config.yaml +++ b/config/config.yaml @@ -17,8 +17,11 @@ summarization: enable: true method: "batch" llm: - model: nvidia/nemotron-3-nano-30b-a3b + model: nvidia/nemotron-3.5-lightning-30b-a3b base_url: https://integrate.api.nvidia.com/v1 + extra_body: + chat_template_kwargs: + thinking: false max_tokens: 2048 temperature: 0.2 top_p: 0.7 @@ -49,8 +52,11 @@ chat: top_k: 25 confidence_threshold: 0.0 llm: - model: nvidia/nemotron-3-nano-30b-a3b + model: nvidia/nemotron-3.5-lightning-30b-a3b base_url: https://integrate.api.nvidia.com/v1 + extra_body: + chat_template_kwargs: + thinking: false max_tokens: 2048 temperature: 0.5 embedding: @@ -64,8 +70,11 @@ notification: enable: false endpoint: "http://127.0.0.1:60000/via-alert-callback" llm: - model: nvidia/nemotron-3-nano-30b-a3b + model: nvidia/nemotron-3.5-lightning-30b-a3b base_url: https://integrate.api.nvidia.com/v1 + extra_body: + chat_template_kwargs: + thinking: false max_tokens: 2048 temperature: 0.2 top_p: 0.7