[v1] Add PrefixLM support to FlexAttention backend (#27938)

Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
This commit is contained in:
Isotr0py
2025-12-07 15:51:36 +00:00
committed by GitHub
parent 541a2ef892
commit b952f4d3c3
16 changed files with 173 additions and 25 deletions
@@ -382,7 +382,6 @@ VLM_TEST_SETTINGS = {
auto_cls=AutoModelForImageTextToText,
vllm_runner_kwargs={"mm_processor_kwargs": {"do_pan_and_scan": True}},
patch_hf_runner=model_utils.gemma3_patch_hf_runner,
num_logprobs=10,
),
"glm4v": VLMTestInfo(
models=["zai-org/glm-4v-9b"],
@@ -30,5 +30,6 @@ class DummyPlatform(Platform):
use_mla,
has_sink,
use_sparse,
use_mm_prefix,
):
return "vllm_add_dummy_platform.dummy_attention_backend.DummyAttentionBackend" # noqa E501