Refactor duplicate torch operation registrations to use shared module

Instead of changing library names (not scalable), create a shared test_operations.py module that:
- Provides a single "silly" library for all compilation tests
- Registers a unified attention operation that can handle both standard and counting modes
- Eliminates duplicate registration errors when running all tests together
- Maintains backward compatibility with existing test behavior

Addresses feedback to make the solution more scalable and maintainable.

Co-authored-by: ProExpertProg <11367180+ProExpertProg@users.noreply.github.com>

tests/compile/piecewise/test_multiple_graphs.py

@@ -6,7 +6,6 @@ are compiled and graph captured separately.
 """
 import torch
 from torch import nn
-from torch.library import Library
 
 from vllm.compilation.backends import set_model_tag
 from vllm.compilation.counter import compilation_counter
@@ -16,10 +15,10 @@ from vllm.config import (CompilationConfig, CompilationLevel, VllmConfig,
                          set_current_vllm_config)
 from vllm.envs import VLLM_USE_V1
 from vllm.forward_context import set_forward_context
-from vllm.utils import direct_register_custom_op
 
-# create a library to hold the custom op
-silly_lib = Library("silly_multiple", "FRAGMENT")  # noqa
+# Import shared test operations
+# The standard attention operation is automatically registered when imported
+import tests.compile.test_operations
 
 BATCH_SIZE = 32
 MLP_SIZE = 128
@@ -27,27 +26,6 @@ HIDDEN_SIZE = 1024
 RANDOM_SEED = 0
 
 
-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,
-)
-
-
 @support_torch_compile
 class ParentModel(nn.Module):
 
@@ -90,7 +68,7 @@ class Attention(nn.Module):
         x = self.pre_attn(x)
         x = self.rms_norm_ref(x)
         attn_output = torch.empty_like(x)
-        torch.ops.silly_multiple.attention(x, x, x, attn_output)
+        torch.ops.silly.attention(x, x, x, attn_output)
         x = attn_output
         x = self.rms_norm_ref(x)
         x = self.post_attn(x)
@@ -188,7 +166,7 @@ def test_ignore_torch_compile_decorator():
         def forward(self, x: torch.Tensor) -> torch.Tensor:
             x = x + x
             attn_output = torch.empty_like(x)
-            torch.ops.silly_multiple.attention(x, x, x, attn_output)
+            torch.ops.silly.attention(x, x, x, attn_output)
             x = attn_output
             x = x * 3
             return x
@@ -347,4 +325,4 @@ def test_multi_graph_piecewise_compile_outputs_equal():
     assert torch.allclose(outputs[0], outputs[1])
 
     # Expect bitwise equivalence using inductor w/ and w/o cudagraph
-    assert torch.equal(outputs[0], outputs[2])
+    assert torch.equal(outputs[0], outputs[2])

tests/compile/piecewise/test_simple.py

@@ -7,7 +7,6 @@ can exactly calculate the expected output and side effects.
 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
@@ -15,36 +14,15 @@ from vllm.config import (CompilationConfig, CompilationLevel, CUDAGraphMode,
                          VllmConfig, set_current_vllm_config)
 from vllm.envs import VLLM_USE_V1
 from vllm.forward_context import BatchDescriptor, set_forward_context
-from vllm.utils import direct_register_custom_op
 
-global_counter = 0
-
-# create a library to hold the custom op
-silly_lib = Library("silly_simple", "FRAGMENT")  # noqa
-
-
-def silly_attention(q: torch.Tensor, k: torch.Tensor, v: torch.Tensor,
-                    out: torch.Tensor) -> None:
-    global global_counter
-    global_counter += 1
-    print(f"{global_counter=}")
-    out.copy_(q)
-    out[0] += 1
-
-
-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,
+# Import shared test operations  
+from tests.compile.test_operations import (
+    get_global_counter, reset_global_counter, enable_counting_mode
 )
 
+# Enable counting mode for this test's specific behavior
+enable_counting_mode()
+
 
 @support_torch_compile
 class SillyModel(nn.Module):
@@ -66,12 +44,12 @@ class SillyModel(nn.Module):
         x = x + 1
         x = x + 2
         out = torch.empty_like(x)
-        torch.ops.silly_simple.attention(x, x, x, out)
+        torch.ops.silly.attention(x, x, x, out)
         x = out
         x = x - 2
         x = x - 1
         out = torch.empty_like(x)
-        torch.ops.silly_simple.attention(x, x, x, out)
+        torch.ops.silly.attention(x, x, x, out)
         x = out
         x = x + 1
         return x
@@ -121,13 +99,12 @@ def test_simple_piecewise_compile(use_inductor):
             model(torch.randn(1).cuda())
 
         input = torch.zeros(2).cuda()
-        global global_counter
-        global_counter = 0
+        reset_global_counter()
         with set_forward_context(
                 None,
                 vllm_config=vllm_config,
                 cudagraph_runtime_mode=CUDAGraphMode.PIECEWISE,
                 batch_descriptor=BatchDescriptor(num_tokens=2, )):
             output = model(input)
-        assert global_counter == 2
-        assert torch.allclose(output.cpu(), torch.tensor([3., 1.]))
+        assert get_global_counter() == 2
+        assert torch.allclose(output.cpu(), torch.tensor([3., 1.]))

tests/compile/piecewise/test_toy_llama.py

@@ -14,38 +14,16 @@ 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_toy_llama", "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,
-)
+# Import shared test operations
+# The standard attention operation is automatically registered when imported
+import tests.compile.test_operations
 
 
 @dataclass
@@ -160,7 +138,7 @@ class LlamaAttention(nn.Module):
         k = k + positions.unsqueeze(1)
 
         attn_output = torch.empty_like(q)
-        torch.ops.silly_toy_llama.attention(q, k, v, attn_output)
+        torch.ops.silly.attention(q, k, v, attn_output)
 
         output = self.output_projection(attn_output)
         return output
@@ -492,4 +470,4 @@ def benchmark():
 
 
 if __name__ == "__main__":
-    benchmark()
+    benchmark()

tests/compile/test_operations.py

@@ -0,0 +1,88 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""
+Shared PyTorch custom operations for compilation tests.
+
+This module provides a centralized place to define and register custom
+PyTorch operations used across multiple compilation tests. This avoids
+duplicate operation registrations that would cause RuntimeErrors when
+running tests together.
+
+The main "attention" operation is automatically registered when this module
+is imported. Individual test files can access additional functionality
+through helper functions.
+"""
+
+import torch
+from torch.library import Library
+
+from vllm.utils import direct_register_custom_op
+
+# Shared library for all compilation test operations
+# Using "silly" namespace to match existing test expectations
+silly_lib = Library("silly", "FRAGMENT")
+
+
+# Global state for test_simple.py compatibility
+_global_counter = 0
+_use_counting_mode = False
+
+
+def get_global_counter():
+    """Get the current global counter value (for test_simple.py)"""
+    return _global_counter
+
+
+def reset_global_counter():
+    """Reset the global counter to 0 (for test_simple.py)"""
+    global _global_counter
+    _global_counter = 0
+
+
+def enable_counting_mode():
+    """Enable counting mode for test_simple.py"""
+    global _use_counting_mode
+    _use_counting_mode = True
+    reset_global_counter()
+
+
+def disable_counting_mode():
+    """Disable counting mode"""
+    global _use_counting_mode
+    _use_counting_mode = False
+
+
+def silly_attention(q: torch.Tensor, k: torch.Tensor, v: torch.Tensor,
+                   out: torch.Tensor) -> None:
+    """
+    Unified attention implementation that can handle both standard and counting modes.
+    """
+    global _global_counter, _use_counting_mode
+    
+    if _use_counting_mode:
+        # Counting mode for test_simple.py
+        _global_counter += 1
+        print(f"global_counter={_global_counter}")
+        out.copy_(q)
+        out[0] += 1
+    else:
+        # Standard mode for test_multiple_graphs.py and test_toy_llama.py
+        out.copy_(q)
+        out += k
+        out += v
+
+
+def silly_attention_fake(q: torch.Tensor, k: torch.Tensor, v: torch.Tensor,
+                        out: torch.Tensor) -> None:
+    """Fake implementation for testing"""
+    return
+
+
+# Register the unified attention operation
+direct_register_custom_op(
+    op_name="attention",
+    op_func=silly_attention,
+    mutates_args=["out"],
+    fake_impl=silly_attention_fake,
+    target_lib=silly_lib,
+)