Route the missing parameter for trtllm_fp8_per_tensor_scale_moe_op

.github/workflows/release.yml

@@ -33,7 +33,7 @@ jobs:
       - name: Validate tag format
         if: github.event_name == 'workflow_dispatch'
         run: |
-          if [[ ! "${{ inputs.tag }}" =~ ^v[0-9]+\.[0-9]+\.[0-9]+([a-z0-9]+)?$ ]]; then
+          if [[ ! "${{ inputs.tag }}" =~ ^v[0-9]+\.[0-9]+\.[0-9]+(\.?[a-z][a-z0-9]*)?$ ]]; then
             echo "Error: Tag '${{ inputs.tag }}' does not match the expected format (e.g., v1.2.3 or v1.2.3.post1 or v1.2.3rc1)"
             exit 1
           fi

flashinfer/fused_moe/core.py

@@ -1240,6 +1240,7 @@ def forward(
                         [-1, -1] if tactic == -1 else tactic,
                         self.activation_type,
                         kwargs.get("norm_topk_prob", True),
+                        kwargs.get("routing_replay_out"),
                     )
             elif (
                 self.dtype_act == DtypeTrtllmGen.Bfloat16
@@ -1625,6 +1626,7 @@ def trtllm_fp8_per_tensor_scale_moe_op(
             [-1, -1] if tactic == -1 else tactic,
             activation_type,
             norm_topk_prob,
+            routing_replay_out,
         )
         if do_finalize:
             return [output]