biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Bugfix] Standardize getting number of image patches/tokens (#34358)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-02-13 12:47:01 +08:00
committed by GitHub
parent 6afa587d31
commit 372b2e762a
29 changed files with 319 additions and 331 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
tests/kernels/core/test_mrope.py
View File

@@ -4,8 +4,6 @@ from typing import NamedTuple
import pytest
import torch
from packaging.version import Version
from transformers import __version__ as TRANSFORMERS_VERSION
from vllm.model_executor.layers.rotary_embedding import get_rope
from vllm.platforms import current_platform
@@ -46,31 +44,13 @@ class MRoPETestInfo(NamedTuple):
marks: list[pytest.MarkDecorator] = []
TRANSFORMERS_BASE_VERSION = Version(TRANSFORMERS_VERSION).base_version
MODELS_TO_TEST = [
MRoPETestInfo(model_name="zai-org/GLM-4.1V-9B-Thinking"),
MRoPETestInfo(model_name="Qwen/Qwen2-VL-7B-Instruct"),
MRoPETestInfo(model_name="Qwen/Qwen2-VL-72B-Instruct"),
MRoPETestInfo(model_name="Qwen/Qwen2.5-VL-72B-Instruct"),
MRoPETestInfo(
model_name="Qwen/Qwen3-VL-4B-Instruct",
marks=[
pytest.mark.skipif(
Version(TRANSFORMERS_BASE_VERSION) < Version("4.57.0"),
reason="Qwen3-VL only available after Transformers v4.57",
)
],
),
MRoPETestInfo(
model_name="Qwen/Qwen3-VL-30B-A3B-Instruct",
marks=[
pytest.mark.skipif(
Version(TRANSFORMERS_BASE_VERSION) < Version("4.57.0"),
reason="Qwen3-VL only available after Transformers v4.57",
)
],
),
MRoPETestInfo(model_name="Qwen/Qwen3-VL-4B-Instruct"),
MRoPETestInfo(model_name="Qwen/Qwen3-VL-30B-A3B-Instruct"),
]
num_tokens_list = [11, 8192]