wip: SE fused SwiGLU deinterleave fix

dsv4/layers/shared_expert.py

@@ -27,6 +27,7 @@ from dsv4.ops.quantize import (
 from dsv4.ops.layouts import (
     make_b_k_major,
     interleave_l1_weights,
+    deinterleave_l1_weights,
 )
 from dsv4.ops.gemm_runner import (
     run_nvfp4_grouped_gemm,
@@ -284,7 +285,11 @@ class Nvfp4SharedExpert:
             global_scale_b=self._l1_gsb,
             swiglu_limit=self.swiglu_limit if self.swiglu_limit is not None else 0.0,
         )
-        return l1_out[:num_tokens]  # (num_tokens, intermediate_size) BF16, SwiGLU already applied
+        l1_out_real = l1_out[:num_tokens]  # (num_tokens, 2*intermediate) BF16, interleaved [silu(gate), silu(gate)*up]
+        # Deinterleave to separate gate and up, then take up half (SwiGLU result)
+        l1_deil = deinterleave_l1_weights(l1_out_real.unsqueeze(0).contiguous())[0]  # (num_tokens, 2*intermediate) deinterleaved
+        intermediate = l1_deil[:, self.intermediate_size:]  # up half = silu(gate)*up
+        return intermediate  # (num_tokens, intermediate_size) BF16
 
     def _run_l1(self, hidden_states: torch.Tensor) -> torch.Tensor:
         """L1 GEMM: activation × gate_up_weight → BF16."""