update sgl_moe_align_block_size usage (#2617)

python/sglang/srt/layers/moe/fused_moe_triton/fused_moe.py

@@ -272,8 +272,14 @@ def moe_align_block_size(
         (max_num_m_blocks,), dtype=torch.int32, device=topk_ids.device
     )
     num_tokens_post_pad = torch.empty((1), dtype=torch.int32, device=topk_ids.device)
-    # FIXME(zhyncs)
-    if not_hip and num_experts >= 256:
+    if not_hip and num_experts >= 224:
+        token_cnts_buffer = torch.empty(
+            (num_experts + 1) * num_experts, dtype=torch.int32, device=topk_ids.device
+        )
+        cumsum_buffer = torch.empty(
+            num_experts + 1, dtype=torch.int32, device=topk_ids.device
+        )
+
         sgl_moe_align_block_size(
             topk_ids,
             num_experts,
@@ -281,6 +287,8 @@ def moe_align_block_size(
             sorted_ids,
             expert_ids,
             num_tokens_post_pad,
+            token_cnts_buffer,
+            cumsum_buffer,
         )
     else:
         ops.moe_align_block_size(