Fused SwiGLU epilogue with granularity-8 weight interleave

- Fix interleave_l1_weights: remove //2 bug (g=granularity_bf16 for N-axis)
- Apply L1 weight+SF interleave in runner._ensure_stacked() and moe_pipeline
- De-interleave L1 GEMM output before gate/up split
- Fused SwiGLU kernel: epi_tile=(128,8) for subtile-level pairing
  - Even subtiles = gate: SiLU in FP32 registers, save to register buffer
  - Odd subtiles = up: silu(gate)*up from buffer
  - Both branches produce same BF16 tensor type (CuTeDSL constraint)
- run_nvfp4_moe_fused() pipeline: fused L1 + PyTorch L2
- Runner: fused_swiglu=True option for CuTeDSLMoERunner
- Layertest: both fused and non-fused paths PASS (cosine 0.988)
- README.md updated with current status and lessons learned

tests/layertest.py

@@ -17,6 +17,7 @@ sys.path.insert(0, REPO_ROOT)
 
 from cutedsl.moe_pipeline import (
     run_nvfp4_moe,
+    run_nvfp4_moe_fused,
 )
 
 NVFP4_MODEL_DIR = "/root/nvidia-meeting/DeepSeek-V4-Pro-NVFP4"
@@ -251,6 +252,32 @@ def main():
     else:
         print(f"  PASS: cosine {cosine:.6f} >= {COSINE_THRESHOLD}")
 
+    # ══════════════════════════════════════════════════════════════
+    # Fused SwiGLU pipeline test
+    # ══════════════════════════════════════════════════════════════
+    print(f"\n  Running CuTeDSL NVFP4 MoE FUSED pipeline (first run compiles)...")
+    fused_output = run_nvfp4_moe_fused(
+        hidden_states, expert_ids, expert_weights,
+        weights, expert_indices,
+    )
+    print(f"    Fused:   amax={fused_output.abs().max():.4f} mean={fused_output.float().mean():.6f}")
+
+    fused_cosine = torch.nn.functional.cosine_similarity(
+        fused_output.flatten().unsqueeze(0).float(),
+        ref_output.flatten().unsqueeze(0).float(),
+    ).item()
+    fused_mse = (fused_output.float() - ref_output.float()).pow(2).mean().item()
+
+    print(f"\n{'=' * 70}")
+    print(f"  FUSED RESULT: cosine={fused_cosine:.6f}  MSE={fused_mse:.6e}")
+    print(f"{'=' * 70}")
+
+    if fused_cosine < COSINE_THRESHOLD:
+        print(f"  FAIL: fused cosine {fused_cosine:.6f} < {COSINE_THRESHOLD}")
+        sys.exit(1)
+    else:
+        print(f"  PASS: fused cosine {fused_cosine:.6f} >= {COSINE_THRESHOLD}")
+
 
 if __name__ == "__main__":
     main()