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 15, 16:00 UTC
[CI] Set max transformers version for Ultravox model test (#16149)
Signed-off-by: Roger Wang <ywang@roblox.com>
Roger Wang committed
Apr 6, 2025 at 21:37 UTC
bb8dab821e00ffdc1007b8b11edbcab6bcfa4140
1 file changed
+2
-1
tests/models/registry.py
+2
-1
@@ -329,7 +329,8 @@ _MULTIMODAL_EXAMPLE_MODELS = {
329
min_transformers_version="4.49"), # noqa: E501
330
"SkyworkR1VChatModel": _HfExamplesInfo("Skywork/Skywork-R1V-38B"),
331
"UltravoxModel": _HfExamplesInfo("fixie-ai/ultravox-v0_5-llama-3_2-1b", # noqa: E501
332
- trust_remote_code=True),
332
+ trust_remote_code=True,
333
+ max_transformers_version="4.50"),
334
# [Encoder-decoder]
335
# Florence-2 uses BartFastTokenizer which can't be loaded from AutoTokenizer
336
# Therefore, we borrow the BartTokenizer from the original Bart model