From e6e054d86e9d9babbd033af80d6d638483243780 Mon Sep 17 00:00:00 2001
From: Cyrus Leung <tlleungac@connect.ust.hk>
Date: Wed, 11 Dec 2024 13:02:02 +0800
Subject: [PATCH] [CI/Build] Check transformers v4.47 (#10991)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
---
 requirements-test.txt                                    | 4 ++--
 .../vision_language/mm_processor_kwargs/test_idefics3.py | 9 ---------
 .../models/embedding/vision_language/test_llava_next.py  | 2 +-
 3 files changed, 3 insertions(+), 12 deletions(-)

diff --git a/requirements-test.txt b/requirements-test.txt
index 38a064bca449a..8ceb705cdffd7 100644
--- a/requirements-test.txt
+++ b/requirements-test.txt
@@ -506,7 +506,7 @@ tiktoken==0.7.0
     #   mistral-common
 timm==1.0.11
     # via -r requirements-test.in
-tokenizers==0.20.3
+tokenizers==0.21.0
     # via transformers
 torch==2.5.1
     # via
@@ -534,7 +534,7 @@ tqdm==4.66.6
     #   transformers
 tqdm-multiprocess==0.0.11
     # via lm-eval
-transformers==4.46.3
+transformers==4.47.0
     # via
     #   lm-eval
     #   peft
diff --git a/tests/models/decoder_only/vision_language/mm_processor_kwargs/test_idefics3.py b/tests/models/decoder_only/vision_language/mm_processor_kwargs/test_idefics3.py
index 31896bfd13e8c..c71a2d359043d 100644
--- a/tests/models/decoder_only/vision_language/mm_processor_kwargs/test_idefics3.py
+++ b/tests/models/decoder_only/vision_language/mm_processor_kwargs/test_idefics3.py
@@ -3,7 +3,6 @@
 
 import pytest
 import torch
-import transformers
 from transformers import AutoImageProcessor, AutoTokenizer
 
 from vllm.inputs import InputContext, token_inputs
@@ -36,8 +35,6 @@ def get_max_idefics3_image_tokens():
     return get_max_idefics3_image_tokens
 
 
-@pytest.mark.skipif(transformers.__version__ < "4.46.0",
-                    reason="Model introduced in HF >= 4.46.0")
 @pytest.mark.parametrize("model", models)
 @pytest.mark.parametrize("longest_edge", [None, 168, 336, 400, 2 * 336])
 def test_input_mapper_override(model: str, image_assets: _ImageAssets,
@@ -77,8 +74,6 @@ def test_input_mapper_override(model: str, image_assets: _ImageAssets,
     assert torch.all(hf_result["pixel_values"] == vllm_result["pixel_values"])
 
 
-@pytest.mark.skipif(transformers.__version__ < "4.46.0",
-                    reason="Model introduced in HF >= 4.46.0")
 @pytest.mark.parametrize("model", models)
 @pytest.mark.parametrize("longest_edge, expected_max_tokens", [
     (None, 2873),
@@ -107,8 +102,6 @@ def test_max_tokens_override(get_max_idefics3_image_tokens, model: str,
     assert expected_max_tokens == actual_max_tokens
 
 
-@pytest.mark.skipif(transformers.__version__ < "4.46.0",
-                    reason="Model introduced in HF >= 4.46.0")
 @pytest.mark.parametrize("model", models)
 @pytest.mark.parametrize("longest_edge, toks_per_img, num_imgs", [
     (168, 169, 1),
@@ -143,8 +136,6 @@ def test_dummy_data_override(dummy_data_for_idefics3, model: str,
     assert img_tok_count == toks_per_img * num_imgs
 
 
-@pytest.mark.skipif(transformers.__version__ < "4.46.0",
-                    reason="Model introduced in HF >= 4.46.0")
 @pytest.mark.parametrize("model", models)
 @pytest.mark.parametrize("longest_edge,expected_toks_per_img,num_imgs", [
     (336, 169 * (1**2 + 1), 1),
diff --git a/tests/models/embedding/vision_language/test_llava_next.py b/tests/models/embedding/vision_language/test_llava_next.py
index 329c6ba279f89..693abd7252d5e 100644
--- a/tests/models/embedding/vision_language/test_llava_next.py
+++ b/tests/models/embedding/vision_language/test_llava_next.py
@@ -86,7 +86,7 @@ def _run_test(
     )
 
 
-@pytest.mark.skipif(transformers.__version__.startswith("4.46"),
+@pytest.mark.skipif(transformers.__version__ >= "4.46",
                     reason="Model broken with changes in transformers 4.46")
 @pytest.mark.core_model
 @pytest.mark.parametrize("model", MODELS)