[NPU][1/N] NPU basic functions refactor and new modelslim quant type (#13359)
This commit is contained in:
@@ -384,7 +384,7 @@ class HiCacheController:
|
||||
self.tp_size = get_tensor_model_parallel_world_size()
|
||||
self.dp_rank = 0
|
||||
|
||||
# Currently, AscendMLAPagedTokenToKVPool is the subclass of MLATokenToKVPool.
|
||||
# Currently, NPUMLATokenToKVPool is the subclass of MLATokenToKVPool.
|
||||
is_mla_backend = isinstance(self.mem_pool_device, MLATokenToKVPool)
|
||||
|
||||
return HiCacheStorageConfig(
|
||||
|
||||
Reference in New Issue
Block a user