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

[CI] And PPL test for Qwen3.5. (#35853)

Browse Source 
Signed-off-by: wang.yuqi <yuqi.wang@daocloud.io>
Signed-off-by: wang.yuqi <noooop@126.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
This commit is contained in:
wang.yuqi
2026-03-03 21:15:51 +08:00
committed by GitHub
parent ad9d09e2b8
commit fd4a90f337
3 changed files with 16 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
tests/models/language/generation_ppl_test/test_qwen.py
View File

@@ -8,14 +8,20 @@ from tests.models.utils import GenerateModelInfo
from .ppl_utils import wikitext_ppl_test
MODELS = [
GenerateModelInfo("Qwen/Qwen3-0.6B"),
GenerateModelInfo("Qwen/Qwen3-0.6B-FP8"),
# transformers:
# Loading a GPTQ quantized model requires optimum, gptqmodel
# GenerateModelInfo("Qwen/Qwen3-0.6B-GPTQ-Int8"),
# for Qwen3
GenerateModelInfo("Qwen/Qwen3-0.6B", hf_ppl=23.864173889160156),
GenerateModelInfo("Qwen/Qwen3-0.6B-FP8", hf_ppl=24.313045501708984),
# for Qwen3.5
GenerateModelInfo("Qwen/Qwen3.5-0.8B", hf_ppl=19.38858413696289),
]
@pytest.mark.parametrize("model_info", MODELS)
def test_ppl(hf_runner, vllm_runner, model_info: GenerateModelInfo):
wikitext_ppl_test(hf_runner, vllm_runner, model_info)
vllm_extra_kwargs = {}
if model_info.name == "Qwen/Qwen3.5-0.8B":
vllm_extra_kwargs["language_model_only"] = True
wikitext_ppl_test(
hf_runner, vllm_runner, model_info, vllm_extra_kwargs=vllm_extra_kwargs
)