[V1][Core] Autotune encoder cache budget (#11895)

Signed-off-by: Roger Wang <ywang@roblox.com>
2025-01-15 11:29:00 -08:00
parent edce722eaa
commit 70755e819e
6 changed files with 166 additions and 49 deletions
--- a/vllm/v1/engine/core.py
+++ b/vllm/v1/engine/core.py
@@ -54,9 +54,12 @@ class EngineCore:
        vllm_config.cache_config.num_cpu_blocks = num_cpu_blocks

        # Setup scheduler.
-        self.scheduler = Scheduler(vllm_config.scheduler_config,
-                                   vllm_config.cache_config,
-                                   vllm_config.lora_config)
+        self.scheduler = Scheduler(
+            scheduler_config=vllm_config.scheduler_config,
+            model_config=vllm_config.model_config,
+            cache_config=vllm_config.cache_config,
+            lora_config=vllm_config.lora_config,
+        )

        self.mm_input_mapper_server = MMInputMapperServer(
            vllm_config.model_config)