Dllm radix cache npu (#27144)
Co-authored-by: ronnie_zheng <zl19940307@163.com>
This commit is contained in:
co-authored by
ronnie_zheng
parent
efbe67d237
commit
a9270250c3
@@ -611,6 +611,7 @@ class Scheduler(
|
|||||||
f"Page size now falls back to {self.dllm_config.block_size}"
|
f"Page size now falls back to {self.dllm_config.block_size}"
|
||||||
)
|
)
|
||||||
self.page_size = self.dllm_config.block_size
|
self.page_size = self.dllm_config.block_size
|
||||||
|
self.server_args.page_size = self.dllm_config.block_size
|
||||||
|
|
||||||
def init_ipc_channels(self, port_args: PortArgs):
|
def init_ipc_channels(self, port_args: PortArgs):
|
||||||
is_rank_zero = (
|
is_rank_zero = (
|
||||||
|
|||||||
Reference in New Issue
Block a user