fix: use BlockSizeTokens param for estimated prefix cache

threcc · threcc · commit b72030080f81 · 2026-03-25T16:58:30.000+01:00
Signed-off-by: threcc &lt;trecchiu@redhat.com&gt;
diff --git a/tests/model_serving/model_server/llmd/llmd_configs/config_estimated_prefix_cache.py b/tests/model_serving/model_server/llmd/llmd_configs/config_estimated_prefix_cache.py
@@ -37,7 +37,7 @@ def _scheduler_config(cls):
                 {
                     "type": "prefix-cache-scorer",
                     "parameters": {
-                        "blockSize": cls.block_size,
+                        "BlockSizeTokens": cls.block_size,
                         "maxPrefixBlocksToMatch": 256,
                         "lruCapacityPerServer": 31250,
                     },

Original file line number	Diff line number	Diff line change
`@@ -37,7 +37,7 @@ def _scheduler_config(cls):`
`37`	`37`	`{`
`38`	`38`	`"type": "prefix-cache-scorer",`
`39`	`39`	`"parameters": {`
`40`		`- "blockSize": cls.block_size,`
	`40`	`+ "BlockSizeTokens": cls.block_size,`
`41`	`41`	`"maxPrefixBlocksToMatch": 256,`
`42`	`42`	`"lruCapacityPerServer": 31250,`
`43`	`43`	`},`