disable overlap schedule if mamba radix cache open (#13057)
This commit is contained in:
@@ -1070,6 +1070,13 @@ class ServerArgs:
|
||||
f"Disabling Radix Cache for {model_arch} as it is not yet supported."
|
||||
)
|
||||
self.disable_radix_cache = True
|
||||
elif model_arch in ["Qwen3NextForCausalLM"]:
|
||||
if not self.disable_radix_cache:
|
||||
logger.warning(
|
||||
"Disabling overlap schedule since MambaRadixCache is not compatible with "
|
||||
"overlap schedule currently, try to use --disable-radix-cache if overlap schedule is necessary"
|
||||
)
|
||||
self.disable_overlap_schedule = True
|
||||
|
||||
if is_deepseek_nsa(hf_config):
|
||||
if (
|
||||
|
||||
Reference in New Issue
Block a user