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 14, 05:30 UTC
[CI/Build] Update transformers to v4.55.2 (#23093)
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Isotr0py committed
Aug 20, 2025 at 01:06 UTC
d6a1a20973dfffcb5eea376e8c19d9659b88aa3c
4 files changed
+12
-11
requirements/common.txt
+1
-1
@@ -7,7 +7,7 @@ requests >= 2.26.0
7
tqdm
8
blake3
9
py-cpuinfo
10
-transformers >= 4.55.0
10
+transformers >= 4.55.2
11
tokenizers >= 0.21.1 # Required for fast incremental detokenization.
12
protobuf # Required by LlamaTokenizer.
13
fastapi[standard] >= 0.115.0 # Required by FastAPI's form models in the OpenAI API server's audio transcriptions endpoint.
requirements/test.in
+1
-1
@@ -34,7 +34,7 @@ opencv-python-headless >= 4.11.0 # required for video test
34
datamodel_code_generator # required for minicpm3 test
35
lm-eval[api]==0.4.8 # required for model evaluation test
36
mteb[bm25s]>=1.38.11, <2 # required for mteb test
37
-transformers==4.55.0
37
+transformers==4.55.2
38
tokenizers==0.21.1
39
schemathesis>=3.39.15 # Required for openai schema test.
40
# quantization
requirements/test.txt
+1
-1
@@ -1139,7 +1139,7 @@ tqdm==4.66.6
1139
# transformers
1140
tqdm-multiprocess==0.0.11
1141
# via lm-eval
1142
-transformers==4.55.0
1142
+transformers==4.55.2
1143
# via
1144
# -r requirements/test.in
1145
# genai-perf
tests/models/multimodal/generation/test_mllama.py
+9
-8
@@ -5,6 +5,7 @@ from typing import Optional, overload
5
6
import pytest
7
import torch
8
+from packaging.version import Version
9
from transformers import AutoConfig, AutoModelForImageTextToText, AutoTokenizer
10
from transformers import __version__ as TRANSFORMERS_VERSION
11
@@ -287,8 +288,8 @@ def clear_cache():
288
@pytest.mark.parametrize("num_logprobs", [5])
289
@pytest.mark.parametrize("attn_backend", LIST_ENC_DEC_SUPPORTED_BACKENDS)
290
@pytest.mark.skipif(
290
- TRANSFORMERS_VERSION == "4.55.0",
291
- reason="Transformers v4.55.0 has a regression issue on mllama, "
291
+ Version(TRANSFORMERS_VERSION) <= Version("4.55.2"),
292
+ reason="Transformers v4.55 has a regression issue on mllama, "
293
"see: https://github.com/huggingface/transformers/pull/40083")
294
def test_models_single_leading_image(hf_runner, vllm_runner, image_assets,
295
model, sizes, dtype, max_tokens,
@@ -319,8 +320,8 @@ def test_models_single_leading_image(hf_runner, vllm_runner, image_assets,
320
@pytest.mark.parametrize("num_logprobs", [5])
321
@pytest.mark.parametrize("attn_backend", LIST_ENC_DEC_SUPPORTED_BACKENDS)
322
@pytest.mark.skipif(
322
- TRANSFORMERS_VERSION == "4.55.0",
323
- reason="Transformers v4.55.0 has a regression issue on mllama, "
323
+ Version(TRANSFORMERS_VERSION) <= Version("4.55.2"),
324
+ reason="Transformers v4.55 has a regression issue on mllama, "
325
"see: https://github.com/huggingface/transformers/pull/40083")
326
def test_models_multi_leading_images(hf_runner, vllm_runner, image_assets,
327
model, dtype, max_tokens, num_logprobs,
@@ -372,8 +373,8 @@ def test_models_multi_leading_images(hf_runner, vllm_runner, image_assets,
373
@pytest.mark.parametrize("num_logprobs", [5])
374
@pytest.mark.parametrize("attn_backend", LIST_ENC_DEC_SUPPORTED_BACKENDS)
375
@pytest.mark.skipif(
375
- TRANSFORMERS_VERSION == "4.55.0",
376
- reason="Transformers v4.55.0 has a regression issue on mllama, "
376
+ Version(TRANSFORMERS_VERSION) <= Version("4.55.2"),
377
+ reason="Transformers v4.55 has a regression issue on mllama, "
378
"see: https://github.com/huggingface/transformers/pull/40083")
379
def test_models_interleaved_images(hf_runner, vllm_runner, image_assets, model,
380
dtype, max_tokens, num_logprobs,
@@ -416,8 +417,8 @@ def test_models_interleaved_images(hf_runner, vllm_runner, image_assets, model,
417
@pytest.mark.parametrize("max_tokens", [64])
418
@pytest.mark.parametrize("num_logprobs", [5])
419
@pytest.mark.skipif(
419
- TRANSFORMERS_VERSION == "4.55.0",
420
- reason="Transformers v4.55.0 has a regression issue on mllama, "
420
+ Version(TRANSFORMERS_VERSION) <= Version("4.55.2"),
421
+ reason="Transformers v4.55 has a regression issue on mllama, "
422
"see: https://github.com/huggingface/transformers/pull/40083")
423
def test_models_distributed(
424
hf_runner,