Skip to content

[llm_bench] Add Qwen3-Omni support (MoE and dense models) #4112

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
avasenin-14 wants to merge 7 commits into
base: master
Choose a base branch
from
+66 −2
Open

[llm_bench] Add Qwen3-Omni support (MoE and dense models) #4112

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
782ba07
Add class for Qwen3-Omni
avasenin-14 Jul 8, 2026
8602664
Add tiny random tests
avasenin-14 Jul 8, 2026
50be29f
Update transformers version check
avasenin-14 Jul 8, 2026
9f227cb
Merge master into the branch
avasenin-14 Jul 8, 2026
7874f92
Add dedicated test in CI
avasenin-14 Jul 8, 2026
375f972
Move the test to tools run
avasenin-14 Jul 8, 2026
712b54f
Add check for other optimum-intel versions
avasenin-14 Jul 9, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions tests/python_tests/samples/conftest.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -197,6 +197,10 @@
"name": "optimum-intel-internal-testing/tiny-random-qwen3-vl",
"convert_args": ["--trust-remote-code", "--task", "image-text-to-text"],
},
"tiny-random-qwen3-omni": {
"name": "optimum-intel-internal-testing/tiny-random-qwen3-omni",
"convert_args": ["--trust-remote-code", "--task", "image-text-to-text"],
},
}

TEST_FILES = {
Expand Down
32 changes: 32 additions & 0 deletions tests/python_tests/samples/test_tools_llm_benchmark.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,13 +4,17 @@
import os
import pytest
import sys
from importlib import metadata
from pathlib import Path

from test_utils import run_sample
from data.models import GGUF_MODEL_LIST
from utils.hugging_face import download_gguf_model
from conftest import SAMPLES_PY_DIR, convert_model, download_test_content


_TRANSFORMERS_VERSION = tuple(int(p) for p in metadata.version("transformers").split(".")[:2])
Comment thread
avasenin-14 marked this conversation as resolved.
Outdated

convert_draft_model = convert_model
download_mask_image = download_test_content

Expand Down Expand Up @@ -615,3 +619,31 @@ def test_python_tool_llm_benchmark_video_gen_json(
generate_llm_bench_input_generation_jsonl,
] + sample_args
run_sample(benchmark_py_command)

@pytest.mark.samples
@pytest.mark.skipif(
_TRANSFORMERS_VERSION < (5, 1),
reason="Requires transformers >= 5.1",
)
Comment thread
avasenin-14 marked this conversation as resolved.
@pytest.mark.parametrize("download_test_content", ["cat.png"], indirect=True)
@pytest.mark.parametrize("convert_model", ["tiny-random-qwen3-omni"], indirect=True)
@pytest.mark.parametrize(
"sample_args",
[
["-d", "cpu", "-n", "1", "--num_steps", "4", "--task", "visual_text_gen", "--optimum"],
["-d", "cpu", "-n", "1", "--num_steps", "4", "--task", "visual_text_gen", "--genai"],
],
)
Comment on lines +727 to +732
def test_python_tool_llm_benchmark_visual_text_gen(self, download_test_content, convert_model, sample_args):
benchmark_script = SAMPLES_PY_DIR / "llm_bench/benchmark.py"
benchmark_py_command = [
sys.executable,
benchmark_script,
"-m",
convert_model,
"--media",
download_test_content,
"--prompt",
"What animal is in this image?",
] + sample_args
run_sample(benchmark_py_command)
4 changes: 3 additions & 1 deletion tools/llm_bench/llm_bench_utils/config_class.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,7 @@
OVDiffusionPipeline,
OVModelForSpeechSeq2Seq,
OVModelForVisualCausalLM,
OVModelForMultimodalLM,
OVPipelineForInpainting,
OVPipelineForImage2Image,
OVModelForFeatureExtraction,
Expand Down Expand Up @@ -171,7 +172,8 @@ class UseCaseTextToSpeech(UseCase):
"videochat-flash-qwen",
"gemma4",
]
)
),
UseCaseVLM(["qwen3-omni"], ov_cls=OVModelForMultimodalLM),
Comment thread
avasenin-14 marked this conversation as resolved.
Outdated
],
"speech_to_text": [UseCaseSpeech2Text(["whisper"])],
"image_cls": [UseCaseImageCls(["vit"])],
Expand Down
Loading