[Bugfix][Model] Honor Molmo2 dummy video num_frames >= 2 override (#53830)
Signed-off-by: hungh <hungh@nvidia.com>
Hung Hoang (黄兴） committed
Aug 26, 2026 at 13:47 UTC
a5f82882a6afa08e65260704011de3321a3aa4a4
2 files changed
+36
-1
tests/models/multimodal/processing/test_molmo2.py
+35
-1
@@ -3,9 +3,14 @@
3
4
from types import SimpleNamespace
5
6
+import pytest
7
import torch
8
8
-from vllm.model_executor.models.molmo2 import build_flat_image_bool_length
9
+from vllm.config.multimodal import VideoDummyOptions
10
+from vllm.model_executor.models.molmo2 import (
11
+ Molmo2DummyInputsBuilder,
12
+ build_flat_image_bool_length,
13
+)
14
15
16
def test_build_flat_image_bool_length_matches_molmoweb_processor_tokens():
@@ -54,3 +59,32 @@ def test_build_flat_image_bool_length_respects_disabled_col_tokens():
59
assert len(image_tokens) == 45
60
assert image_tokens[0].item() == hf_config.low_res_image_start_token_id
61
assert (image_tokens == hf_config.image_col_id).sum().item() == 0
62
+
63
+
64
+@pytest.mark.parametrize(
65
+ ("num_frames_override", "expected_frames"),
66
+ [(1, 2), (2, 2), (3, 3)],
67
+)
68
+def test_dummy_video_num_frames_override_honors_min_of_two(
69
+ num_frames_override, expected_frames
70
+):
71
+ """A ``num_frames`` override below 2 must be ignored (the model needs at
72
+ least 2 frames), matching the "cannot be less than 2, will be ignored"
73
+ warning."""
74
+ builder = object.__new__(Molmo2DummyInputsBuilder)
75
+ builder.info = SimpleNamespace(
76
+ get_hf_processor=lambda: SimpleNamespace(
77
+ video_processor=SimpleNamespace(size={"width": 64, "height": 64})
78
+ ),
79
+ get_num_frames_with_most_features=lambda seq_len, mm_counts: 16,
80
+ get_image_size_with_most_features=lambda: (64, 64),
81
+ )
82
+
83
+ data = builder.get_dummy_mm_data(
84
+ seq_len=128,
85
+ mm_counts={"image": 0, "video": 1},
86
+ mm_options={"video": VideoDummyOptions(num_frames=num_frames_override)},
87
+ )
88
+
89
+ video, _metadata = data["video"][0]
90
+ assert video.shape[0] == expected_frames
vllm/model_executor/models/molmo2.py
+1
@@ -1879,6 +1879,7 @@ class Molmo2DummyInputsBuilder(BaseDummyInputsBuilder[Molmo2ProcessingInfo]):
1879
num_frames_override,
1880
)
1881
target_num_frames = min(target_num_frames, num_frames_override)
1882
+ target_num_frames = max(target_num_frames, 2)
1883
1884
dummy_videos = self._get_dummy_videos(
1885
width=video_size["width"],