A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[BugFix] Also consider RAY_EXPERIMENTAL_NOSET_* when storing compilation cache (#27294)
Signed-off-by: Hollow Man <hollowman@opensuse.org>
ℍ𝕠𝕝𝕝𝕠𝕨 𝕄𝕒𝕟 committed
Oct 28, 2025 at 16:22 UTC
936643a86837db51f05974c71a05a9b89941b014
1 file changed
+23
vllm/envs.py
+23
@@ -1544,6 +1544,29 @@ def compute_hash() -> str:
1544
1545
factors = [environment_variables[key]() for key in environment_variables_to_hash]
1546
1547
+ ray_noset_env_vars = [
1548
+ # Refer to
1549
+ # https://github.com/ray-project/ray/blob/c584b1ea97b00793d1def71eaf81537d70efba42/python/ray/_private/accelerators/nvidia_gpu.py#L11
1550
+ # https://github.com/ray-project/ray/blob/c584b1ea97b00793d1def71eaf81537d70efba42/python/ray/_private/accelerators/amd_gpu.py#L11
1551
+ # https://github.com/ray-project/ray/blob/b97d21dab233c2bd8ed7db749a82a1e594222b5c/python/ray/_private/accelerators/amd_gpu.py#L10
1552
+ # https://github.com/ray-project/ray/blob/c584b1ea97b00793d1def71eaf81537d70efba42/python/ray/_private/accelerators/npu.py#L12
1553
+ # https://github.com/ray-project/ray/blob/c584b1ea97b00793d1def71eaf81537d70efba42/python/ray/_private/accelerators/hpu.py#L12
1554
+ # https://github.com/ray-project/ray/blob/c584b1ea97b00793d1def71eaf81537d70efba42/python/ray/_private/accelerators/neuron.py#L14
1555
+ # https://github.com/ray-project/ray/blob/c584b1ea97b00793d1def71eaf81537d70efba42/python/ray/_private/accelerators/tpu.py#L38
1556
+ # https://github.com/ray-project/ray/blob/c584b1ea97b00793d1def71eaf81537d70efba42/python/ray/_private/accelerators/intel_gpu.py#L10
1557
+ # https://github.com/ray-project/ray/blob/c584b1ea97b00793d1def71eaf81537d70efba42/python/ray/_private/accelerators/rbln.py#L10
1558
+ "RAY_EXPERIMENTAL_NOSET_CUDA_VISIBLE_DEVICES",
1559
+ "RAY_EXPERIMENTAL_NOSET_ROCR_VISIBLE_DEVICES",
1560
+ "RAY_EXPERIMENTAL_NOSET_HIP_VISIBLE_DEVICES",
1561
+ "RAY_EXPERIMENTAL_NOSET_ASCEND_RT_VISIBLE_DEVICES",
1562
+ "RAY_EXPERIMENTAL_NOSET_HABANA_VISIBLE_MODULES",
1563
+ "RAY_EXPERIMENTAL_NOSET_NEURON_RT_VISIBLE_CORES",
1564
+ "RAY_EXPERIMENTAL_NOSET_TPU_VISIBLE_CHIPS",
1565
+ "RAY_EXPERIMENTAL_NOSET_ONEAPI_DEVICE_SELECTOR",
1566
+ "RAY_EXPERIMENTAL_NOSET_RBLN_RT_VISIBLE_DEVICES",
1567
+ ]
1568
+ factors.extend([os.getenv(var) for var in ray_noset_env_vars])
1569
+
1570
hash_str = hashlib.md5(str(factors).encode(), usedforsecurity=False).hexdigest()
1571
1572
return hash_str