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
[V0 Deprecation][TPU] Remove V1 flag check from tests (#22248)
Signed-off-by: NickLucche <nlucches@redhat.com>
Nicolò Lucchesi committed
Aug 5, 2025 at 15:53 UTC
0c275ad5ad1af35636581bffaafc9e694a270378
3 files changed
+1
-21
tests/v1/tpu/test_mha_attn.py
-7
@@ -12,17 +12,10 @@ import torch_xla
12
import torch_xla.core
13
import torch_xla.core.xla_model
14
15
-from vllm import envs
15
from vllm.attention.layer import MultiHeadAttention
16
from vllm.attention.selector import _cached_get_attn_backend
17
from vllm.platforms import current_platform
18
20
-if not envs.VLLM_USE_V1:
21
- pytest.skip(
22
- "Skipping V1 tests. Rerun with `VLLM_USE_V1=1` to test.",
23
- allow_module_level=True,
24
- )
25
-
19
20
@pytest.fixture(autouse=True)
21
def clear_cache():
tests/v1/tpu/test_multimodal.py
-7
@@ -4,19 +4,12 @@
4
import openai
5
import pytest
6
7
-from vllm import envs
7
from vllm.multimodal.utils import encode_image_base64, fetch_image
8
from vllm.platforms import current_platform
9
10
from ...entrypoints.openai.test_vision import TEST_IMAGE_URLS
11
from ...utils import RemoteOpenAIServer
12
14
-if not envs.VLLM_USE_V1:
15
- pytest.skip(
16
- "Skipping V1 tests. Rerun with `VLLM_USE_V1=1` to test.",
17
- allow_module_level=True,
18
- )
19
-
13
14
@pytest.fixture(scope="session")
15
def base64_encoded_image() -> dict[str, str]:
tests/v1/tpu/test_sampler.py
+1
-7
@@ -4,16 +4,10 @@ import random
4
5
import pytest
6
7
-from vllm import LLM, envs
7
+from vllm import LLM
8
from vllm.platforms import current_platform
9
from vllm.sampling_params import SamplingParams
10
11
-if not envs.VLLM_USE_V1:
12
- pytest.skip(
13
- "Skipping V1 tests. Rerun with `VLLM_USE_V1=1` to test.",
14
- allow_module_level=True,
15
- )
16
-
11
12
@pytest.mark.parametrize("model_name", ["Qwen/Qwen2.5-1.5B-Instruct"])
13
@pytest.mark.skipif(not current_platform.is_tpu(),