Skip to content

Commit 5027ed7

Browse files
authored
【BugFif】fix ep decode (#4138)
* suppoort expert num 3 pre rank * fix ep decode
1 parent 25aa2d9 commit 5027ed7

File tree

1 file changed

+1
-1
lines changed

1 file changed

+1
-1
lines changed

fastdeploy/model_executor/layers/moe/fused_moe_backend_base.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -166,7 +166,7 @@ def apply(
166166
else:
167167
if layer.fd_config.parallel_config.splitwise_role == "mixed":
168168
self.ep_decoder_runner.clean_low_latency_buffer()
169-
return self.apply_ep_prefill(layer, x, gate)
169+
return self.apply_ep_decode(layer, x, gate)
170170
else:
171171
return self.apply_tp(layer, x, gate)
172172

0 commit comments

Comments
 (0)