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

[Bugfix] Correct behavior of GraniteMoeHybrid for TensorParallel execution (#20137)

Browse Source 
Signed-off-by: Stanislaw Wozniak <stw@zurich.ibm.com>
This commit is contained in:
Stan Wozniak
2025-06-28 17:16:41 +02:00
committed by GitHub
parent daceac57c7
commit daec9dea6e
3 changed files with 73 additions and 78 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
tests/models/language/generation/test_hybrid.py
View File

@@ -28,8 +28,9 @@ SSM_MODELS = [
HYBRID_MODELS = [
"ai21labs/Jamba-tiny-dev",
# NOTE: ibm-granite/granite-4.0-tiny-preview are skipped currently as
# it is not yet available in huggingface transformers
# NOTE: Currently the test failes due to HF transformers issue fixed in:
# https://github.com/huggingface/transformers/pull/39033
# We will enable vLLM test for Granite after next HF transformers release.
# "ibm-granite/granite-4.0-tiny-preview",
# NOTE: Running Plamo2 in transformers implementation requires to install
# causal-conv1d package, which is not listed as a test dependency as it's