Revert "Fix memory leak when doing chunked prefill" (#1797)
This commit is contained in:
@@ -663,7 +663,6 @@ def run_mmlu_test(
|
||||
chunked_prefill_size=32,
|
||||
):
|
||||
other_args = ["--chunked-prefill-size", str(chunked_prefill_size)]
|
||||
other_args += ["--mem-fraction-static", "0.85"]
|
||||
if disable_radix_cache:
|
||||
other_args += ["--disable-radix-cache"]
|
||||
if enable_mixed_chunk:
|
||||
|
||||
Reference in New Issue
Block a user