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 6, 05:45 UTC
[CI] Skip ISAAC multimodal tests due to broken upstream HF model weights (#37781)
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Andreas Karatzas committed
Mar 22, 2026 at 00:23 UTC
c862481c02e8c9ec3bafee83bb7768997a5d54ee
1 file changed
+5
-1
tests/models/multimodal/generation/test_common.py
+5
-1
@@ -545,8 +545,12 @@ VLM_TEST_SETTINGS = {
545
auto_cls=AutoModelForImageTextToText,
546
),
547
"isaac": VLMTestInfo(
548
+ # NOTE: PerceptronAI/Isaac-0.1 removed because the upstream HF
549
+ # repo has a stale model.safetensors.index.json that references
550
+ # shard files which no longer exist (consolidated into a single
551
+ # model.safetensors on 2026-03-20). Re-add once upstream fixes
552
+ # the index file.
553
models=[
549
- "PerceptronAI/Isaac-0.1",
554
"PerceptronAI/Isaac-0.2-2B-Preview",
555
],
556
test_type=(VLMTestType.IMAGE, VLMTestType.MULTI_IMAGE),