add return_expert_indices=True in ep_prefill

ckl117 · ckl117 · commit 8eb2f1e74134 · 2026-04-10T17:43:25.000+08:00
diff --git a/fastdeploy/model_executor/layers/moe/fused_moe_cutlass_backend.py b/fastdeploy/model_executor/layers/moe/fused_moe_cutlass_backend.py
@@ -168,6 +168,7 @@ def apply_ep_prefill(
                         tokens_per_expert=[],
                         padding_alignment=128,
                         override_buffer_size=token_all_num,
+                        return_expert_indices=True,
                     )
                 )
 

Original file line number	Diff line number	Diff line change
`@@ -168,6 +168,7 @@ def apply_ep_prefill(`
`168`	`168`	`tokens_per_expert=[],`
`169`	`169`	`padding_alignment=128,`
`170`	`170`	`override_buffer_size=token_all_num,`
	`171`	`+ return_expert_indices=True,`
`171`	`172`	`)`
`172`	`173`	`)`
`173`	`174`