@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 444e2e7e1f
Fork 0

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 9, 19:30 UTC

[Misc] Bump opencv-python dependecy version to 4.13 (#32668)

Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>

Isotr0py committed Jan 22, 2026 at 23:51 UTC 444e2e7e1f79f3c8fdf57c362272127df55e4847
5 files changed +35 -16
requirements/common.txt
+1 -1
@@ -32,7 +32,7 @@ pyzmq >= 25.0.0
32 msgspec
33 gguf >= 0.17.0
34 mistral_common[image] >= 1.8.8
35 -opencv-python-headless >= 4.11.0 # required for video IO
35 +opencv-python-headless >= 4.13.0 # required for video IO
36 pyyaml
37 six>=1.16.0; python_version > '3.11' # transitive dependency of pandas that needs to be the latest version for python 3.12
38 setuptools>=77.0.3,<81.0.0; python_version > '3.11' # Setuptools is used by triton, we need to ensure a modern version is installed for 3.12+ so that it does not try to import distutils, which was removed in 3.12
requirements/nightly_torch_test.txt
+3 -3
@@ -25,7 +25,7 @@ transformers_stream_generator # required for qwen-vl test
25 matplotlib # required for qwen-vl test
26 mistral_common[image,audio] >= 1.8.8 # required for voxtral test
27 num2words # required for smolvlm test
28 -opencv-python-headless >= 4.11.0 # required for video test
28 +opencv-python-headless >= 4.13.0 # required for video test
29 datamodel_code_generator # required for minicpm3 test
30 lm-eval[api]>=0.4.9.2 # required for model evaluation test
31 mteb>=1.38.11, <2 # required for mteb test
@@ -37,8 +37,8 @@ bitsandbytes>=0.46.1
37 buildkite-test-collector==0.1.9
38
39
40 -genai_perf==0.0.8
41 -tritonclient==2.51.0
40 +genai_perf>=0.0.8
41 +tritonclient>=2.51.0
42
43 numba == 0.61.2 # Required for N-gram speculative decoding
44 numpy
requirements/test.in
+3 -3
@@ -33,7 +33,7 @@ matplotlib # required for qwen-vl test
33 mistral_common[image,audio] >= 1.8.8 # required for voxtral test
34 num2words # required for smolvlm test
35 open_clip_torch==2.32.0 # Required for nemotron_vl test, Nemotron Parse in test_common.py
36 -opencv-python-headless >= 4.11.0 # required for video test
36 +opencv-python-headless >= 4.13.0 # required for video test
37 datamodel_code_generator # required for minicpm3 test
38 lm-eval[api]>=0.4.9.2 # required for model evaluation test
39 mteb[bm25s]>=2, <3 # required for mteb test
@@ -45,8 +45,8 @@ bitsandbytes==0.46.1
45 buildkite-test-collector==0.1.9
46
47
48 -genai_perf==0.0.8
49 -tritonclient==2.51.0
48 +genai_perf>=0.0.8
49 +tritonclient>=2.51.0
50
51 arctic-inference == 0.1.1 # Required for suffix decoding test
52 numba == 0.61.2 # Required for N-gram speculative decoding
requirements/test.txt
+23 -8
@@ -31,7 +31,9 @@ albumentations==1.4.6
31 # -r requirements/test.in
32 # terratorch
33 alembic==1.16.4
34 - # via mlflow
34 + # via
35 + # mlflow
36 + # optuna
37 annotated-doc==0.0.4
38 # via fastapi
39 annotated-types==0.7.0
@@ -145,6 +147,8 @@ colorama==0.4.6
147 # tqdm-multiprocess
148 colorful==0.5.6
149 # via ray
150 +colorlog==6.10.1
151 + # via optuna
152 contourpy==1.3.0
153 # via matplotlib
154 coverage==7.10.6
@@ -252,7 +256,7 @@ fsspec==2024.9.0
256 # torch
257 ftfy==6.3.1
258 # via open-clip-torch
255 -genai-perf==0.0.8
259 +genai-perf==0.0.16
260 # via -r requirements/test.in
261 genson==1.3.0
262 # via datamodel-code-generator
@@ -389,6 +393,7 @@ jinja2==3.1.6
393 # via
394 # datamodel-code-generator
395 # flask
396 + # genai-perf
397 # mlflow
398 # torch
399 jiwer==3.0.5
@@ -528,7 +533,7 @@ numba==0.61.2
533 # librosa
534 numexpr==2.10.1
535 # via lm-eval
531 -numpy==1.26.4
536 +numpy==2.2.6
537 # via
538 # -r requirements/test.in
539 # accelerate
@@ -558,6 +563,7 @@ numpy==1.26.4
563 # numba
564 # numexpr
565 # opencv-python-headless
566 + # optuna
567 # pandas
568 # patsy
569 # peft
@@ -637,7 +643,7 @@ opencensus==0.11.4
643 # via ray
644 opencensus-context==0.1.3
645 # via opencensus
640 -opencv-python-headless==4.11.0.86
646 +opencv-python-headless==4.13.0.90
647 # via
648 # -r requirements/test.in
649 # albucore
@@ -660,6 +666,10 @@ opentelemetry-sdk==1.35.0
666 # ray
667 opentelemetry-semantic-conventions==0.56b0
668 # via opentelemetry-sdk
669 +optuna==3.6.1
670 + # via genai-perf
671 +orjson==3.11.5
672 + # via genai-perf
673 packaging==24.2
674 # via
675 # accelerate
@@ -678,6 +688,7 @@ packaging==24.2
688 # lightning-utilities
689 # matplotlib
690 # mlflow-skinny
691 + # optuna
692 # peft
693 # plotly
694 # pooch
@@ -717,6 +728,8 @@ peft==0.16.0
728 # lm-eval
729 perceptron==0.1.4
730 # via -r requirements/test.in
731 +perf-analyzer==0.1.0
732 + # via genai-perf
733 pillow==10.4.0
734 # via
735 # genai-perf
@@ -903,6 +916,7 @@ pyyaml==6.0.2
916 # lightning
917 # mlflow-skinny
918 # omegaconf
919 + # optuna
920 # peft
921 # pytorch-lightning
922 # ray
@@ -1065,6 +1079,7 @@ sortedcontainers==2.4.0
1079 soundfile==0.12.1
1080 # via
1081 # -r requirements/test.in
1082 + # genai-perf
1083 # librosa
1084 # mistral-common
1085 soxr==0.5.0.post1
@@ -1075,6 +1090,7 @@ sqlalchemy==2.0.41
1090 # via
1091 # alembic
1092 # mlflow
1093 + # optuna
1094 sqlitedict==2.1.0
1095 # via lm-eval
1096 sqlparse==0.5.3
@@ -1204,6 +1220,7 @@ tqdm==4.66.6
1220 # mteb
1221 # nltk
1222 # open-clip-torch
1223 + # optuna
1224 # peft
1225 # pqdm
1226 # pretrainedmodels
@@ -1226,10 +1243,8 @@ transformers-stream-generator==0.0.5
1243 # via -r requirements/test.in
1244 triton==3.5.1
1245 # via torch
1229 -tritonclient==2.51.0
1230 - # via
1231 - # -r requirements/test.in
1232 - # genai-perf
1246 +tritonclient==2.64.0
1247 + # via -r requirements/test.in
1248 typepy==1.3.2
1249 # via
1250 # dataproperty
tests/entrypoints/openai/test_translation_validation.py
+5 -1
@@ -267,12 +267,16 @@ async def test_audio_with_max_tokens(mary_had_lamb, client_and_model):
267 out_tokens = tok(out_text, add_special_tokens=False)["input_ids"]
268 assert len(out_tokens) == 1
269 # max_completion_tokens > max_model_len
270 + # max_model_len=32768 for Gemma-3n-E2B-it
271 transcription = await client.audio.transcriptions.create(
272 model=model_name,
273 file=mary_had_lamb,
274 response_format="text",
275 temperature=0.0,
275 - extra_body={"max_completion_tokens": int(1e6)},
276 + extra_body={
277 + "max_completion_tokens": int(1e6),
278 + "repetition_penalty": 1.3,
279 + },
280 )
281 out = json.loads(transcription)
282 out_text = out["text"]