[Refactor] Refactor batch invariance folder (#53619)

Signed-off-by: yewentao256 <zhyanwentao@126.com>

.github/CODEOWNERS

@@ -11,7 +11,7 @@
 /vllm/model_executor/layers/mamba @tdoublep @tomeras91
 /vllm/model_executor/layers/mamba/gdn/qwen_gdn_linear_attn.py @tdoublep @ZJY0516 @vadiklyutiy
 /vllm/model_executor/model_loader @22quinn
-/vllm/model_executor/layers/batch_invariant.py @yewentao256
+/vllm/model_executor/determinism/ @yewentao256
 /vllm/model_executor/custom_op.py @shen-shanshan
 /vllm/ir @ProExpertProg
 /vllm/kernels/ @ProExpertProg @tjtanaa

examples/rl/rlhf_async_new_apis.py

@@ -123,7 +123,7 @@ class TrainModel:
     """Ray actor that wraps the training model on a dedicated GPU."""
 
     def __init__(self, model_name: str):
-        from vllm.model_executor.layers.batch_invariant import (
+        from vllm.model_executor.determinism.batch_invariant import (
             init_batch_invariance,
         )
 

tests/v1/determinism/test_matmul_batch_invariant.py

@@ -11,8 +11,8 @@ import pytest
 import torch
 from utils import skip_unsupported
 
-from vllm.model_executor.layers.batch_invariant import matmul_batch_invariant
-from vllm.model_executor.layers.batch_invariant_configs import (
+from vllm.model_executor.determinism.batch_invariant import matmul_batch_invariant
+from vllm.model_executor.determinism.batch_invariant_configs import (
     _BATCH_INVARIANT_MATMUL_TUNED_CONFIGS,
     _get_tuned_matmul_arch_family,
 )

tests/v1/determinism/test_rms_norm_batch_invariant.py

@@ -6,7 +6,7 @@ import pytest
 import torch
 from utils import skip_if_not_cuda, skip_unsupported
 
-from vllm.model_executor.layers.batch_invariant import (
+from vllm.model_executor.determinism.batch_invariant import (
     rms_norm_batch_invariant,
 )
 from vllm.platforms import current_platform

vllm/model_executor/determinism/__init__.py

@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project

vllm/model_executor/determinism/batch_invariant.py

@@ -8,7 +8,7 @@ from typing import Any
 import torch
 
 import vllm.envs as envs
-from vllm.model_executor.layers.batch_invariant_configs import (
+from vllm.model_executor.determinism.batch_invariant_configs import (
     _get_matmul_config,
     resolve_tuned_matmul_configs,
 )

vllm/model_executor/layers/layernorm.py

@@ -11,7 +11,7 @@ import vllm.kernels  # noqa: F401
 from vllm import envs, ir
 from vllm.logger import init_logger
 from vllm.model_executor.custom_op import CustomOp
-from vllm.model_executor.layers.batch_invariant import rms_norm_batch_invariant
+from vllm.model_executor.determinism.batch_invariant import rms_norm_batch_invariant
 
 logger = init_logger(__name__)
 

vllm/model_executor/layers/linear.py

@@ -21,7 +21,7 @@ from vllm.distributed import (
 )
 from vllm.logger import init_logger
 from vllm.model_executor.custom_op import PluggableLayer
-from vllm.model_executor.layers.batch_invariant import (
+from vllm.model_executor.determinism.batch_invariant import (
     linear_batch_invariant,
 )
 from vllm.model_executor.layers.quantization.base_config import (

vllm/model_executor/layers/vocab_parallel_embedding.py

@@ -16,7 +16,7 @@ from vllm.distributed import (
     tensor_model_parallel_all_reduce,
 )
 from vllm.model_executor.custom_op import PluggableLayer
-from vllm.model_executor.layers.batch_invariant import (
+from vllm.model_executor.determinism.batch_invariant import (
     linear_batch_invariant,
 )
 from vllm.model_executor.layers.quantization.base_config import (

vllm/v1/worker/gpu_worker.py

@@ -1416,7 +1416,7 @@ def init_worker_distributed_environment(
 ) -> None:
     """Initialize the distributed environment."""
     parallel_config = vllm_config.parallel_config
-    from vllm.model_executor.layers.batch_invariant import init_batch_invariance
+    from vllm.model_executor.determinism.batch_invariant import init_batch_invariance
 
     init_batch_invariance()
     override_envs_for_eplb(