Revert "[CI] Remove deleted nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 and its arch aliases" (#55392)

This commit is contained in:
Misha Goin
2026-09-04 14:45:24 -07:00
committed by GitHub
parent 3f41d102c5
commit 8f269a93bb
3 changed files with 40 additions and 0 deletions
+30
View File
@@ -7,6 +7,7 @@ from typing import Any, Literal
import pytest
from packaging.version import Version
from transformers import PretrainedConfig
from transformers import __version__ as TRANSFORMERS_VERSION
from vllm.config.model import ModelDType, TokenizerMode
@@ -1201,6 +1202,35 @@ _MULTIMODAL_EXAMPLE_MODELS = {
},
trust_remote_code=True,
),
"NemotronH_Nano_Omni_Reasoning_V3": _HfExamplesInfo(
"nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16",
max_model_len=4096,
# NemotronH layers are constructed via `hybrid_override_pattern`
use_original_num_layers=True,
hf_overrides={
"vision_config": PretrainedConfig(
args={
"min_num_patches": 1,
"max_num_patches": 12,
"model": "vit_huge_patch16_224",
},
video_temporal_patch_size=2,
# TODO(nhaber): This is `true` in the official `config.json`,
# but this causes a processor exception in the tests due to a known bug
# with mixed-resolution video when `true`. To be resolved.
video_maintain_aspect_ratio=False,
),
"text_config": {"num_hidden_layers": 2, "hybrid_override_pattern": "M*"},
},
trust_remote_code=True,
),
"NemotronH_Super_Omni_Reasoning_V3": _HfExamplesInfo(
"nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16", is_available_online=False
),
# TODO: Change repo id once pertinent archs are public.
"NemotronH_Omni_Reasoning_V3": _HfExamplesInfo(
"nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16", is_available_online=False
),
"OpenCUAForConditionalGeneration": _HfExamplesInfo(
"xlangai/OpenCUA-7B",
trust_remote_code=True,
+7
View File
@@ -797,6 +797,13 @@ class SpeculativeConfig:
{"n_predict": n_predict, "architectures": ["ErnieMTPModel"]}
)
if hf_config.architectures[0] in (
"NemotronH_Super_Omni_Reasoning_V3",
"NemotronH_Omni_Reasoning_V3",
):
# Promote VLM's text_config so MTP detection below fires correctly
hf_config = hf_config.text_config
if (
hf_config.model_type in {"nemotron_h", "nemotron_h_puzzle"}
and hasattr(hf_config, "num_nextn_predict_layers")
+3
View File
@@ -520,6 +520,9 @@ _MULTIMODAL_MODELS = {
"MossAudioModel": ("moss_audio", "MossAudioModel"),
"HfMoondream": ("moondream3", "Moondream3ForCausalLM"),
"NemotronH_Nano_VL_V2": ("nano_nemotron_vl", "NemotronH_Nano_VL_V2"),
"NemotronH_Nano_Omni_Reasoning_V3": ("nano_nemotron_vl", "NemotronH_Nano_VL_V2"),
"NemotronH_Super_Omni_Reasoning_V3": ("nano_nemotron_vl", "NemotronH_Nano_VL_V2"),
"NemotronH_Omni_Reasoning_V3": ("nano_nemotron_vl", "NemotronH_Nano_VL_V2"),
"NVLM_D": ("nvlm_d", "NVLM_D_Model"),
"MuseGlimmerForConditionalGeneration": ("muse_glimmer", "MuseGlimmerForCausalLM"),
"OpenCUAForConditionalGeneration": ("opencua", "OpenCUAForConditionalGeneration"),