[CI][ROCm] Deprecate DinD for MI250 test groups (#56162)

Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
This commit is contained in:
Andreas Karatzas
2026-09-18 14:13:35 -07:00
committed by GitHub
co-authored by OpenAI Codex
parent 49e0427ca5
commit 2c3fb4c545
18 changed files with 102 additions and 6 deletions
+1 -1
View File
@@ -241,7 +241,7 @@ def dist_init():
rank=0,
distributed_init_method=f"file://{temp_file}",
local_rank=0,
backend="nccl",
backend="gloo" if current_platform.is_cpu() else "nccl",
)
initialize_model_parallel(1, 1)
yield
@@ -0,0 +1,38 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from functools import partial
import pytest
import torch
import torch.nn as nn
from vllm.model_executor.layers.activation import GeluAndMul, SiluAndMul
from vllm.platforms import CpuArchEnum, current_platform
@pytest.mark.skipif(
not current_platform.is_cpu()
or current_platform.get_cpu_architecture() == CpuArchEnum.POWERPC,
reason="Native CPU activations require a non-PowerPC CPU platform",
)
@pytest.mark.parametrize(
"activation_factory,reference",
[
(SiluAndMul, nn.SiLU()),
(GeluAndMul, nn.GELU()),
(partial(GeluAndMul, approximate="tanh"), nn.GELU(approximate="tanh")),
],
ids=["silu", "gelu", "gelu_tanh"],
)
def test_cpu_act_and_mul_without_compiled_ops(
default_vllm_config, monkeypatch, activation_factory, reference
):
"""CPU activations must construct and run without compiled wheel operators."""
default_vllm_config.compilation_config.custom_ops = ["all"]
monkeypatch.setattr(torch.ops, "_C", object())
x = torch.tensor([[-2.0, -1.0, 0.0, 1.0, 2.0, 3.0, 4.0, 5.0]], device="cpu")
layer = activation_factory()
torch.testing.assert_close(layer(x), reference(x[:, :4]) * x[:, 4:])
@@ -155,10 +155,15 @@ def test_vllm_openai_image_embeds_metadata_contract() -> None:
def test_rust_build_cache_excludes_git_metadata() -> None:
import torch
from vllm.platforms import current_platform
dockerfile_names = ["Dockerfile", "Dockerfile.cpu"]
if not current_platform.is_rocm():
# CPU jobs can reuse ROCm artifacts, which omit the XPU Dockerfile.
if not current_platform.is_rocm() and (
not current_platform.is_cpu() or torch.version.hip is None
):
dockerfile_names.append("Dockerfile.xpu")
for name in dockerfile_names:
dockerfile = (REPO_ROOT / "docker" / name).read_text()