[Bugfix] Fix inconsistent handling of cache reset (#33481)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
@@ -717,6 +717,10 @@ class GPUModelRunner(
|
||||
self.effective_drafter_max_model_len = self.max_model_len
|
||||
|
||||
def reset_mm_cache(self) -> None:
|
||||
"""
|
||||
Clear the multi-modal cache that was used during profiling,
|
||||
but no longer needed during inference.
|
||||
"""
|
||||
if self.mm_budget:
|
||||
self.mm_budget.reset_cache()
|
||||
|
||||
|
||||
Reference in New Issue
Block a user