[CI] Remove deleted nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 and its arch aliases (#55026)
Signed-off-by: khluu <khluu000@gmail.com> Co-authored-by: Kimi Code <noreply@moonshot.cn>
Kevin H. Luu committed
Sep 2, 2026 at 15:18 UTC
6c6376a09e89acfbe62954246739f0dd52b892c7
3 files changed
-40
tests/models/registry.py
-30
@@ -7,7 +7,6 @@ from typing import Any, Literal
7
8
import pytest
9
from packaging.version import Version
10
-from transformers import PretrainedConfig
10
from transformers import __version__ as TRANSFORMERS_VERSION
11
12
from vllm.config.model import ModelDType, TokenizerMode
@@ -1195,35 +1194,6 @@ _MULTIMODAL_EXAMPLE_MODELS = {
1194
},
1195
trust_remote_code=True,
1196
),
1198
- "NemotronH_Nano_Omni_Reasoning_V3": _HfExamplesInfo(
1199
- "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16",
1200
- max_model_len=4096,
1201
- # NemotronH layers are constructed via `hybrid_override_pattern`
1202
- use_original_num_layers=True,
1203
- hf_overrides={
1204
- "vision_config": PretrainedConfig(
1205
- args={
1206
- "min_num_patches": 1,
1207
- "max_num_patches": 12,
1208
- "model": "vit_huge_patch16_224",
1209
- },
1210
- video_temporal_patch_size=2,
1211
- # TODO(nhaber): This is `true` in the official `config.json`,
1212
- # but this causes a processor exception in the tests due to a known bug
1213
- # with mixed-resolution video when `true`. To be resolved.
1214
- video_maintain_aspect_ratio=False,
1215
- ),
1216
- "text_config": {"num_hidden_layers": 2, "hybrid_override_pattern": "M*"},
1217
- },
1218
- trust_remote_code=True,
1219
- ),
1220
- "NemotronH_Super_Omni_Reasoning_V3": _HfExamplesInfo(
1221
- "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16", is_available_online=False
1222
- ),
1223
- # TODO: Change repo id once pertinent archs are public.
1224
- "NemotronH_Omni_Reasoning_V3": _HfExamplesInfo(
1225
- "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16", is_available_online=False
1226
- ),
1197
"OpenCUAForConditionalGeneration": _HfExamplesInfo(
1198
"xlangai/OpenCUA-7B",
1199
trust_remote_code=True,
vllm/config/speculative.py
-7
@@ -796,13 +796,6 @@ class SpeculativeConfig:
796
{"n_predict": n_predict, "architectures": ["ErnieMTPModel"]}
797
)
798
799
- if hf_config.architectures[0] in (
800
- "NemotronH_Super_Omni_Reasoning_V3",
801
- "NemotronH_Omni_Reasoning_V3",
802
- ):
803
- # Promote VLM's text_config so MTP detection below fires correctly
804
- hf_config = hf_config.text_config
805
-
799
if (
800
hf_config.model_type in {"nemotron_h", "nemotron_h_puzzle"}
801
and hasattr(hf_config, "num_nextn_predict_layers")
vllm/model_executor/models/registry.py
-3
@@ -514,9 +514,6 @@ _MULTIMODAL_MODELS = {
514
"MossAudioModel": ("moss_audio", "MossAudioModel"),
515
"HfMoondream": ("moondream3", "Moondream3ForCausalLM"),
516
"NemotronH_Nano_VL_V2": ("nano_nemotron_vl", "NemotronH_Nano_VL_V2"),
517
- "NemotronH_Nano_Omni_Reasoning_V3": ("nano_nemotron_vl", "NemotronH_Nano_VL_V2"),
518
- "NemotronH_Super_Omni_Reasoning_V3": ("nano_nemotron_vl", "NemotronH_Nano_VL_V2"),
519
- "NemotronH_Omni_Reasoning_V3": ("nano_nemotron_vl", "NemotronH_Nano_VL_V2"),
517
"NVLM_D": ("nvlm_d", "NVLM_D_Model"),
518
"MuseGlimmerForConditionalGeneration": ("muse_glimmer", "MuseGlimmerForCausalLM"),
519
"OpenCUAForConditionalGeneration": ("opencua", "OpenCUAForConditionalGeneration"),