[CI] execute all piecewise compilation tests together (#24502)

Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>

tests/compile/piecewise/test_toy_llama.py

@@ -14,38 +14,15 @@ from typing import Any, Optional
 import pytest
 import torch
 from torch import nn
-from torch.library import Library
 
 from vllm.compilation.counter import compilation_counter
 from vllm.compilation.decorators import support_torch_compile
 from vllm.config import (CompilationConfig, CompilationLevel, CUDAGraphMode,
                          VllmConfig, set_current_vllm_config)
 from vllm.forward_context import BatchDescriptor, set_forward_context
-from vllm.utils import direct_register_custom_op
 
-# create a library to hold the custom op
-silly_lib = Library("silly", "FRAGMENT")  # noqa
-
-
-def silly_attention(q: torch.Tensor, k: torch.Tensor, v: torch.Tensor,
-                    out: torch.Tensor) -> None:
-    out.copy_(q)
-    out += k
-    out += v
-
-
-def silly_attention_fake(q: torch.Tensor, k: torch.Tensor, v: torch.Tensor,
-                         out: torch.Tensor) -> None:
-    return
-
-
-direct_register_custom_op(
-    op_name="attention",
-    op_func=silly_attention,
-    mutates_args=["out"],
-    fake_impl=silly_attention_fake,
-    target_lib=silly_lib,
-)
+# This import automatically registers `torch.ops.silly.attention`
+from .. import silly_attention  # noqa: F401
 
 
 @dataclass