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

[Model] New model support for Motif-1-Tiny (#23414)

Browse Source 
Signed-off-by: ca1207 <ca1207zzz@gmail.com>
Signed-off-by: TaehyunKim <73943231+ca1207@users.noreply.github.com>
Co-authored-by: WyldeCat <skan1543@gmail.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
This commit is contained in:
TaehyunKim
2025-09-11 15:29:40 +09:00
committed by GitHub
parent e2b1f863aa
commit 9bd831f501
13 changed files with 871 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
tests/models/test_initialization.py
View File

@@ -63,8 +63,9 @@ def can_initialize(model_arch: str, monkeypatch: pytest.MonkeyPatch,
_initialize_kv_caches_v1), monkeypatch.context() as m):
if model_info.v0_only:
m.setenv("VLLM_USE_V1", "0")
if model_arch == "Phi4FlashForCausalLM":
# Phi4FlashForCausalLM only supports DIFFERENTIAL_FLASH_ATTN backend
if model_arch in ("Phi4FlashForCausalLM", "MotifForCausalLM"):
# Phi4FlashForCausalLM and MotifForCausalLM
# only supports DIFFERENTIAL_FLASH_ATTN backend
m.setenv("VLLM_ATTENTION_BACKEND", "DIFFERENTIAL_FLASH_ATTN")
if model_arch == "GptOssForCausalLM":
# FIXME: A hack to bypass FA3 assertion because our CI's L4 GPU