[CI] Upgrade trasnformers version (#6307)

Upgrade transformers to >=4.56.4

- vLLM version: v0.14.1
- vLLM main:
dc917cceb8

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>

tests/e2e/singlecard/test_aclgraph_mem.py

@@ -21,7 +21,6 @@ from unittest.mock import patch
 
 import pytest
 import torch
-from modelscope import snapshot_download  # type: ignore
 from vllm import SamplingParams
 
 from tests.e2e.conftest import VllmRunner
@@ -66,11 +65,11 @@ def test_aclgraph_mem_use(model: str, max_tokens: int) -> None:
         sampling_params = SamplingParams(max_tokens=max_tokens,
                                          temperature=0.0)
         if model == "vllm-ascend/DeepSeek-V2-Lite-W8A8":
-            vllm_model = VllmRunner(snapshot_download(model),
+            vllm_model = VllmRunner(model,
                                     max_model_len=1024,
                                     quantization="ascend")
         else:
-            vllm_model = VllmRunner(snapshot_download(model))
+            vllm_model = VllmRunner(model)
         _ = vllm_model.generate(prompts, sampling_params)
 
     assert capture_called.value == 1, "capture_model was not called during test"

tests/e2e/singlecard/test_ilama_lora.py

@@ -1,6 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
 import vllm
-from modelscope import snapshot_download  # type: ignore
 from vllm.lora.request import LoRARequest
 
 from tests.e2e.conftest import VllmRunner
@@ -46,7 +45,7 @@ def do_sample(llm: vllm.LLM, lora_path: str, lora_id: int) -> list[str]:
 
 def test_ilama_lora(ilama_lora_files):
     with VllmRunner(
-            snapshot_download(MODEL_PATH),
+            MODEL_PATH,
             enable_lora=True,
             dtype="half",
             max_loras=4,

tests/e2e/singlecard/test_models.py

@@ -20,7 +20,6 @@
 import os
 
 import pytest
-from modelscope import snapshot_download  # type: ignore
 from vllm import SamplingParams
 from vllm.assets.audio import AudioAsset
 
@@ -46,7 +45,7 @@ def test_minicpm(model) -> None:
     ]
     max_tokens = 5
 
-    with VllmRunner(snapshot_download(model),
+    with VllmRunner(model,
                     max_model_len=512,
                     gpu_memory_utilization=0.7) as runner:
         runner.generate_greedy(example_prompts, max_tokens)
@@ -61,7 +60,7 @@ def test_whisper(model) -> None:
                                      max_tokens=10,
                                      stop_token_ids=None)
 
-    with VllmRunner(snapshot_download(model),
+    with VllmRunner(model,
                     max_model_len=448,
                     max_num_seqs=5,
                     dtype="bfloat16",

tests/e2e/singlecard/test_quantization.py

@@ -15,8 +15,6 @@
 # limitations under the License.
 # This file is a part of the vllm-ascend project.
 #
-from modelscope import snapshot_download  # type: ignore[import-untyped]
-
 from tests.e2e.conftest import VllmRunner
 from tests.e2e.model_utils import check_outputs_equal
 
@@ -33,7 +31,7 @@ def test_qwen3_w8a8_quant():
                             )]
 
     with VllmRunner(
-            snapshot_download("vllm-ascend/Qwen3-0.6B-W8A8"),
+            "vllm-ascend/Qwen3-0.6B-W8A8",
             max_model_len=8192,
             gpu_memory_utilization=0.7,
             cudagraph_capture_sizes=[1, 2, 4, 8],
@@ -62,7 +60,7 @@ def test_qwen3_dense_w8a16():
                             )]
 
     with VllmRunner(
-            snapshot_download("vllm-ascend/Qwen3-0.6B-W8A16"),
+            "vllm-ascend/Qwen3-0.6B-W8A16",
             max_model_len=8192,
             enforce_eager=False,
             gpu_memory_utilization=0.7,