[MLX] Fix startup after logical token capacity change (#41314)

This commit is contained in:
Wenfei (Charles) Qi
2026-09-29 21:15:56 -07:00
committed by GitHub
parent 9ee065e49a
commit 7fb85a9d88
2 changed files with 44 additions and 0 deletions
@@ -107,6 +107,13 @@ class MlxModelRunnerStub(ModelRunner):
# that path working instead of raising AttributeError.
prefill_aware_swa = False
@property
def logical_max_total_num_tokens(self) -> int:
"""MLX pool slots already count logical tokens; no DCP expansion applies."""
return self.req_to_token_pool.schedulable_token_capacity(
self.max_total_num_tokens
)
@property
def preloaded_weights_bytes(self) -> int:
"""Return zero for the base Torch loader accounting hook.