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] Fix failing multimodal standard test (#22153)
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Isotr0py committed
Aug 4, 2025 at 03:04 UTC
6a39ba85fe0f2fff9494b5eccea717c93510c230
2 files changed
+10
-4
tests/models/multimodal/test_tensor_schema.py
+2
@@ -105,6 +105,8 @@ def test_model_tensor_schema(model_arch: str, vllm_runner: type[VllmRunner],
105
106
model_info = HF_EXAMPLE_MODELS.get_hf_info(model_arch)
107
model_info.check_available_online(on_fail="skip")
108
+ model_info.check_transformers_version(on_fail="skip",
109
+ check_max_version=False)
110
111
model_id = model_info.default
112
tests/models/registry.py
+8
-4
@@ -80,6 +80,8 @@ class _HfExamplesInfo:
80
self,
81
*,
82
on_fail: Literal["error", "skip"],
83
+ check_min_version: bool = True,
84
+ check_max_version: bool = True,
85
) -> None:
86
"""
87
If the installed transformers version does not meet the requirements,
@@ -96,9 +98,11 @@ class _HfExamplesInfo:
98
msg = f"`transformers=={current_version}` installed, but `transformers"
99
# Only check the base version for the min/max version, otherwise preview
100
# models cannot be run because `x.yy.0.dev0`<`x.yy.0`
99
- if min_version and Version(cur_base_version) < Version(min_version):
101
+ if (check_min_version and min_version
102
+ and Version(cur_base_version) < Version(min_version)):
103
msg += f">={min_version}` is required to run this model."
101
- elif max_version and Version(cur_base_version) > Version(max_version):
104
+ elif (check_max_version and max_version
105
+ and Version(cur_base_version) > Version(max_version)):
106
msg += f"<={max_version}` is required to run this model."
107
else:
108
return
@@ -185,6 +189,8 @@ _TEXT_GENERATION_EXAMPLE_MODELS = {
189
min_transformers_version="4.53"),
190
"GlmForCausalLM": _HfExamplesInfo("THUDM/glm-4-9b-chat-hf"),
191
"Glm4ForCausalLM": _HfExamplesInfo("THUDM/GLM-4-9B-0414"),
192
+ "Glm4MoeForCausalLM": _HfExamplesInfo("zai-org/GLM-4.5",
193
+ min_transformers_version="4.54"), # noqa: E501
194
"GPT2LMHeadModel": _HfExamplesInfo("openai-community/gpt2",
195
{"alias": "gpt2"}),
196
"GPTBigCodeForCausalLM": _HfExamplesInfo("bigcode/starcoder",
@@ -378,8 +384,6 @@ _MULTIMODAL_EXAMPLE_MODELS = {
384
trust_remote_code=True,
385
hf_overrides={"architectures": ["GLM4VForCausalLM"]}), # noqa: E501
386
"Glm4vForConditionalGeneration": _HfExamplesInfo("THUDM/GLM-4.1V-9B-Thinking"), # noqa: E501
381
- "Glm4MoeForCausalLM": _HfExamplesInfo("zai-org/GLM-4.5",
382
- min_transformers_version="4.54"), # noqa: E501
387
"Glm4v_moeForConditionalGeneration": _HfExamplesInfo("zai-org/GLM-4.5V-Air",
388
is_available_online=False), # noqa: E501
389
"H2OVLChatModel": _HfExamplesInfo("h2oai/h2ovl-mississippi-800m",