[CPU] Avoid prefill CP predicates during decode graph capture (#39690)
Co-authored-by: Ma Mingfei <mingfei.ma@intel.com>
This commit is contained in:
co-authored by
Ma Mingfei
parent
f65c70bb7d
commit
a407915c17
@@ -915,7 +915,10 @@ class LayerCommunicator:
|
||||
return True
|
||||
if forward_batch.dp_padding_mode.is_max_len():
|
||||
return True
|
||||
if dsa_use_prefill_cp(forward_batch) or is_mla_cp_active(forward_batch):
|
||||
# Prefill CP predicates must stay out of decode graph capture.
|
||||
if forward_batch.forward_mode.is_context_parallel_extend() and (
|
||||
dsa_use_prefill_cp(forward_batch) or is_mla_cp_active(forward_batch)
|
||||
):
|
||||
return True
|
||||
if get_attn_tp_context().input_scattered and not self.is_last_layer:
|
||||
return True
|
||||
|
||||
Reference in New Issue
Block a user