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 4, 15:00 UTC
[Tests] Gate Isaac under Transformers v5 (#40907)
Signed-off-by: Silu Panda <31051721+SiluPanda@users.noreply.github.com>
Silu Panda committed
Apr 26, 2026 at 19:26 UTC
c0879d94839a4cc0febba20cc1cb5642fc5c9cc4
1 file changed
+12
tests/models/registry.py
+12
@@ -961,6 +961,18 @@ _MULTIMODAL_EXAMPLE_MODELS = {
961
"PerceptronAI/Isaac-0.1",
962
trust_remote_code=True,
963
extras={"0.2-2B-Preview": "PerceptronAI/Isaac-0.2-2B-Preview"},
964
+ max_transformers_version="4.57",
965
+ transformers_version_reason={
966
+ "vllm": (
967
+ "Custom Isaac code is not compatible with Transformers v5. "
968
+ "The model should be upstreamed to Transformers for "
969
+ "long-term support."
970
+ ),
971
+ "hf": (
972
+ "Isaac's remote model and processor code import or configure "
973
+ "APIs that changed in Transformers v5."
974
+ ),
975
+ },
976
),
977
"InternS1ForConditionalGeneration": _HfExamplesInfo(
978
"internlm/Intern-S1",