Merge pull request #11 from XyLearningProgramming/bugfix/oom

XyLearningProgramming · web-flow · commit a10784af8246 · 2026-02-19T23:20:00.000+08:00
hard limited n ctx to 2048
diff --git a/slm_server/config.py b/slm_server/config.py
@@ -62,7 +62,7 @@ class Settings(BaseSettings):
         description="Owner label for /models list. Set SLM_MODEL_OWNER to override.",
     )
     n_ctx: int = Field(
-        8192, description="Maximum context window (input + generated tokens)."
+        2048, description="Maximum context window (input + generated tokens)."
     )
     n_threads: int = Field(
         2, description="Number of OpenMP threads llama‑cpp will spawn."

Original file line number	Diff line number	Diff line change
`@@ -62,7 +62,7 @@ class Settings(BaseSettings):`
`62`	`62`	`description="Owner label for /models list. Set SLM_MODEL_OWNER to override.",`
`63`	`63`	`)`
`64`	`64`	`n_ctx: int = Field(`
`65`		`- 8192, description="Maximum context window (input + generated tokens)."`
	`65`	`+ 2048, description="Maximum context window (input + generated tokens)."`
`66`	`66`	`)`
`67`	`67`	`n_threads: int = Field(`
`68`	`68`	`2, description="Number of OpenMP threads llama‑cpp will spawn."`