remove redundant maybe_calc_kv_scales

vllm/attention/layer.py

@@ -432,43 +432,6 @@ def maybe_calc_kv_scales_fake(    query: torch.Tensor,
     return
 
 
-direct_register_custom_op(
-    op_name="maybe_calc_kv_scales",
-    op_func=maybe_calc_kv_scales,
-    mutates_args=[],
-    fake_impl=maybe_calc_kv_scales_fake,
-    dispatch_key=current_platform.dispatch_key,
-    tags=tag_cudagraph_unsafe,)
-
-def maybe_calc_kv_scales(
-    query: torch.Tensor,
-    key: torch.Tensor,
-    value: torch.Tensor,
-    layer_name: str,
-) -> None:
-
-    forward_context: ForwardContext = get_forward_context()
-    attn_metadata = forward_context.attn_metadata
-
-    if isinstance(attn_metadata, dict):
-        attn_metadata = attn_metadata[layer_name]
-
-    # if attn_metadata is None or not getattr(
-    #         attn_metadata, 'enable_kv_scales_calculation', False):
-    #     return
-
-    self = forward_context.no_compile_layers[layer_name]
-    self.calc_kv_scales(query, key, value)
-
-
-def maybe_calc_kv_scales_fake(    query: torch.Tensor,
-    key: torch.Tensor,
-    value: torch.Tensor,
-    layer_name: str,
-) -> None:
-    return
-
-
 direct_register_custom_op(
     op_name="maybe_calc_kv_scales",
     op_func=maybe_calc_kv_scales,