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, 20:45 UTC
Upgrade transformers-4.57.5 (#32287)
Signed-off-by: Huy Do <huydhn@gmail.com>
Huy Do committed
Jan 21, 2026 at 21:19 UTC
f5fdec8ce21ae97d1140613b7a265a1d6af844f0
4 files changed
+25
-3
requirements/nightly_torch_test.txt
+1
-1
@@ -29,7 +29,7 @@ opencv-python-headless >= 4.11.0 # 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
32
-transformers==4.57.3
32
+transformers==4.57.5
33
tokenizers==0.22.0
34
schemathesis>=3.39.15 # Required for openai schema test.
35
# quantization
requirements/test.in
+1
-1
@@ -37,7 +37,7 @@ opencv-python-headless >= 4.11.0 # 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
40
-transformers==4.57.3
40
+transformers==4.57.5
41
tokenizers==0.22.0
42
schemathesis>=3.39.15 # Required for openai schema test.
43
# quantization
requirements/test.txt
+1
-1
@@ -1214,7 +1214,7 @@ tqdm==4.66.6
1214
# transformers
1215
tqdm-multiprocess==0.0.11
1216
# via lm-eval
1217
-transformers==4.57.3
1217
+transformers==4.57.5
1218
# via
1219
# -r requirements/test.in
1220
# genai-perf
tests/models/multimodal/pooling/test_jinavl_reranker.py
+22
@@ -3,6 +3,8 @@
3
from typing import cast
4
5
import pytest
6
+import transformers
7
+from packaging import version
8
from transformers import AutoModel
9
10
from vllm.entrypoints.chat_utils import (
@@ -277,6 +279,10 @@ def _run_test(
279
280
@pytest.mark.parametrize("model", MODELS)
281
@pytest.mark.parametrize("dtype", ["half"])
282
+@pytest.mark.skipif(
283
+ version.parse(transformers.__version__) == version.parse("4.57.5"),
284
+ reason="Skipped for transformers==4.57.5, https://github.com/huggingface/transformers/issues/43295",
285
+)
286
def test_model_text_image(
287
hf_runner,
288
vllm_runner,
@@ -296,6 +302,10 @@ def test_model_text_image(
302
303
@pytest.mark.parametrize("model", MODELS)
304
@pytest.mark.parametrize("dtype", ["half"])
305
+@pytest.mark.skipif(
306
+ version.parse(transformers.__version__) == version.parse("4.57.5"),
307
+ reason="Skipped for transformers==4.57.5, https://github.com/huggingface/transformers/issues/43295",
308
+)
309
def test_model_text_text(
310
hf_runner,
311
vllm_runner,
@@ -315,6 +325,10 @@ def test_model_text_text(
325
326
@pytest.mark.parametrize("model", MODELS)
327
@pytest.mark.parametrize("dtype", ["half"])
328
+@pytest.mark.skipif(
329
+ version.parse(transformers.__version__) == version.parse("4.57.5"),
330
+ reason="Skipped for transformers==4.57.5, https://github.com/huggingface/transformers/issues/43295",
331
+)
332
def test_model_image_text(
333
hf_runner,
334
vllm_runner,
@@ -334,6 +348,10 @@ def test_model_image_text(
348
349
@pytest.mark.parametrize("model", MODELS)
350
@pytest.mark.parametrize("dtype", ["half"])
351
+@pytest.mark.skipif(
352
+ version.parse(transformers.__version__) == version.parse("4.57.5"),
353
+ reason="Skipped for transformers==4.57.5, https://github.com/huggingface/transformers/issues/43295",
354
+)
355
def test_model_image_image(
356
hf_runner,
357
vllm_runner,
@@ -353,6 +371,10 @@ def test_model_image_image(
371
372
@pytest.mark.parametrize("model", MODELS)
373
@pytest.mark.parametrize("dtype", ["half"])
374
+@pytest.mark.skipif(
375
+ version.parse(transformers.__version__) == version.parse("4.57.5"),
376
+ reason="Skipped for transformers==4.57.5, https://github.com/huggingface/transformers/issues/43295",
377
+)
378
def test_model_text_mixed_documents(
379
hf_runner,
380
vllm_runner,