Update refactoring operation

vllm/attention/ops/paged_attn.py

@@ -193,28 +193,28 @@ class PagedAttention:
                 
             if envs.VLLM_USE_OPT_OP:
                 ops.paged_attention_v2_opt(
-                output,
-                exp_sums,
-                max_logits,
-                tmp_output,
-                query,
-                key_cache,
-                value_cache,
-                num_kv_heads,
-                scale,
-                block_tables,
-                seq_lens,
-                block_size,
-                max_seq_len,
-                alibi_slopes,
-                kv_cache_dtype,
-                kv_scale,
-                tp_rank,
-                blocksparse_local_blocks,
-                blocksparse_vert_stride,
-                blocksparse_block_size,
-                blocksparse_head_sliding_step,
-            )
+                    output,
+                    exp_sums,
+                    max_logits,
+                    tmp_output,
+                    query,
+                    key_cache,
+                    value_cache,
+                    num_kv_heads,
+                    scale,
+                    block_tables,
+                    seq_lens,
+                    block_size,
+                    max_seq_len,
+                    alibi_slopes,
+                    kv_cache_dtype,
+                    kv_scale,
+                    tp_rank,
+                    blocksparse_local_blocks,
+                    blocksparse_vert_stride,
+                    blocksparse_block_size,
+                    blocksparse_head_sliding_step,
+                )
             else:
                 ops.paged_attention_v2(
                     output,

vllm/model_executor/layers/activation.py

@@ -36,7 +36,7 @@ class SiluAndMul(CustomOp):
         output_shape = (x.shape[:-1] + (d, ))
         out = torch.empty(output_shape, dtype=x.dtype, device=x.device)
         if envs.VLLM_USE_OPT_OP:
-            ops.silu_and_mul(out, x)
+            ops.silu_and_mul_opt(out, x)
         else:
             ops.silu_and_mul(out, x)
         return out

vllm/model_executor/layers/layernorm.py

@@ -70,11 +70,11 @@ class RMSNorm(CustomOp):
         out = torch.empty_like(x)
         if envs.VLLM_USE_OPT_OP:
             ops.rms_norm_opt(
-            out,
-            x,
-            self.weight.data,
-            self.variance_epsilon,
-        )
+                out,
+                x,
+                self.weight.data,
+                self.variance_epsilon,
+            )
         else:
             ops.rms_norm(
                 out,