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 3, 17:45 UTC
[CI] set max transformers version for skywork model (#42104)
Signed-off-by: Divakar Verma <divakar.verma@amd.com>
Divakar Verma committed
May 13, 2026 at 19:53 UTC
ca7e4546da4cb716df671dfcf26e459dd3dfd9f0
1 file changed
+10
-1
tests/models/registry.py
+10
-1
@@ -1334,7 +1334,16 @@ _MULTIMODAL_EXAMPLE_MODELS = {
1334
),
1335
"RForConditionalGeneration": _HfExamplesInfo("YannQi/R-4B", trust_remote_code=True),
1336
"SkyworkR1VChatModel": _HfExamplesInfo(
1337
- "Skywork/Skywork-R1V-38B", trust_remote_code=True
1337
+ "Skywork/Skywork-R1V-38B",
1338
+ trust_remote_code=True,
1339
+ max_transformers_version="4.57",
1340
+ transformers_version_reason={
1341
+ "hf": (
1342
+ "SkyworkChatModel.__init__ does not call self.post_init(), so "
1343
+ "`all_tied_weights_keys` is never set; Transformers v5 requires "
1344
+ "this attribute in _move_missing_keys_from_meta_to_device."
1345
+ )
1346
+ },
1347
),
1348
"SmolVLMForConditionalGeneration": _HfExamplesInfo(
1349
"HuggingFaceTB/SmolVLM2-2.2B-Instruct"