GeneralBots/botserver
2
1
Fork 0
Code Issues Pull requests Projects Releases 323 Packages Wiki Activity Actions

fix: Increase default n_predict to 512 for DeepSeek R1 reasoning
All checks were successful
BotServer CI / build (push) Successful in 9m26s
Details

Browse source 
DeepSeek R1 model outputs reasoning_content first, then content.
With n_predict=50, responses were truncated during reasoning phase.
Increased to 512 to allow full reasoning + response.

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
This commit is contained in:
Rodrigo Rodriguez 2026-02-15 20:27:35 +00:00
parent a9cbbbffa0
commit 0a1bd25869
1 changed files with 2 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
src/llm/local.rs
View file

@ -386,8 +386,8 @@ pub fn start_llm_server(
let n_predict = config_manager let n_predict = config_manager
.get_config(&default_bot_id, "llm-server-n-predict", None) .get_config(&default_bot_id, "llm-server-n-predict", None)
.unwrap_or_else(|_| "50".to_string()); .unwrap_or_else(|_| "512".to_string()); // Increased default for DeepSeek R1 reasoning
let n_predict = if n_predict.is_empty() { "50".to_string() } else { n_predict }; let n_predict = if n_predict.is_empty() { "512".to_string() } else { n_predict };
let n_ctx_size = config_manager let n_ctx_size = config_manager
.get_config(&default_bot_id, "llm-server-ctx-size", None) .get_config(&default_bot_id, "llm-server-ctx-size", None)