[CI][Misc] Use offline mode for model downloads (#7179)

### What this PR does / why we need it?
1. For all parts of the current test module involving the millisecond
download model, add the `local_file_only` parameter to specify offline
mode; this ensures that CI will not fail due to network instability.
2. Install modelscope from a fixed commit until it next release
### Does this PR introduce _any_ user-facing change?

### How was this patch tested?
 check if the env or arg `local_files_only` works
1) set the env:
```shell
export HF_HUB_OFFLINE=1
```
2) run the script
```python
from transformers import PretrainedConfig
import huggingface_hub
from modelscope.utils.hf_util import patch_hub

patch_hub()

model="Qwen/Qwen3-0.6B"
kwargs = {}


config_dict, _ = PretrainedConfig.get_config_dict(
    model,
    trust_remote_code=True,
    local_files_only=huggingface_hub.constants.HF_HUB_OFFLINE,
    **kwargs,
)

print(config_dict)
```
it works well:
```shell
2026-03-06 06:40:12,546 - modelscope - WARNING - We can not confirm the cached file is for revision: master
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
{'architectures': ['Qwen3ForCausalLM'], 'attention_bias': False, 'attention_dropout': 0.0, 'bos_token_id': 151643, 'eos_token_id': 151645, 'head_dim': 128, 'hidden_act': 'silu', 'hidden_size': 1024, 'initializer_range': 0.02, 'intermediate_size': 3072, 'max_position_embeddings': 40960, 'max_window_layers': 28, 'model_type': 'qwen3', 'num_attention_heads': 16, 'num_hidden_layers': 28, 'num_key_value_heads': 8, 'rms_norm_eps': 1e-06, 'rope_scaling': None, 'rope_theta': 1000000, 'sliding_window': None, 'tie_word_embeddings': True, 'torch_dtype': 'bfloat16', 'transformers_version': '4.51.0', 'use_cache': True, 'use_sliding_window': False, 'vocab_size': 151936, '_commit_hash': None}
```
3) test the model repo does not cached locally when the env
`HF_HUB_OFFLINE`==True
```python
from transformers import PretrainedConfig
import huggingface_hub
from modelscope.utils.hf_util import patch_hub

patch_hub()


model="FireRedTeam/FireRed-OCR"
kwargs = {}


config_dict, _ = PretrainedConfig.get_config_dict(
    model,
    trust_remote_code=True,
    local_files_only=huggingface_hub.constants.HF_HUB_OFFLINE,
    **kwargs,
)

print(config_dict)
```
and the result is as expected:
```shell
  File "/workspace/demo.py", line 12, in <module>
    config_dict, _ = PretrainedConfig.get_config_dict(
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/python3.11.14/lib/python3.11/site-packages/modelscope/utils/hf_util/patcher.py", line 189, in patch_get_config_dict
    model_dir = get_model_dir(pretrained_model_name_or_path,
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/python3.11.14/lib/python3.11/site-packages/modelscope/utils/hf_util/patcher.py", line 164, in get_model_dir
    model_dir = snapshot_download(
                ^^^^^^^^^^^^^^^^^^
  File "/usr/local/python3.11.14/lib/python3.11/site-packages/modelscope/hub/snapshot_download.py", line 137, in snapshot_download
    return _snapshot_download(
           ^^^^^^^^^^^^^^^^^^^
  File "/usr/local/python3.11.14/lib/python3.11/site-packages/modelscope/hub/snapshot_download.py", line 283, in _snapshot_download
    raise ValueError(
ValueError: Cannot find the requested files in the cached path and outgoing traffic has been disabled. To enable look-ups and downloads online, set 'local_files_only' to False
```
- vLLM version: v0.16.0
- vLLM main:
15d76f74e2

---------

Signed-off-by: wangli <wangli858794774@gmail.com>

tests/e2e/conftest.py

@@ -34,6 +34,7 @@ import traceback
 from pathlib import Path
 from typing import Any, TypeVar
 
+import huggingface_hub
 import numpy as np
 import openai
 import psutil
@@ -1024,7 +1025,10 @@ class HfRunner:
 
 @pytest.fixture(scope="session")
 def ilama_lora_files():
-    return snapshot_download(repo_id="vllm-ascend/ilama-text2sql-spider")
+    return snapshot_download(
+        repo_id="vllm-ascend/ilama-text2sql-spider",
+        local_files_only=huggingface_hub.constants.HF_HUB_OFFLINE,
+    )
 
 
 @pytest.fixture(scope="session")

tests/e2e/multicard/2-cards/test_external_launcher.py

@@ -26,6 +26,7 @@ import sys
 from pathlib import Path
 from unittest.mock import patch
 
+import huggingface_hub
 import pytest
 import torch_npu
 from modelscope import snapshot_download  # type: ignore
@@ -165,7 +166,10 @@ def test_qwen3_external_launcher_with_sleepmode():
 def test_qwen3_external_launcher_with_sleepmode_level2():
     script = Path(__file__).parent.parent.parent.parent.parent / "examples" / "offline_external_launcher.py"
     env = os.environ.copy()
-    model_path = snapshot_download("Qwen/Qwen3-8B")
+    model_path = snapshot_download(
+        "Qwen/Qwen3-8B",
+        local_files_only=huggingface_hub.constants.HF_HUB_OFFLINE,
+    )
     # TODO: Add moe model test
     cmd = [
         sys.executable,

tests/e2e/singlecard/pooling/test_classification.py

@@ -1,13 +1,14 @@
 import torch
 from modelscope import snapshot_download  # type: ignore[import-untyped]
 from transformers import AutoModelForSequenceClassification
+import huggingface_hub
 
 from tests.e2e.conftest import HfRunner, VllmRunner
 
 
 def test_qwen_pooling_classify_correctness() -> None:
 
-    model_name = snapshot_download("Howeee/Qwen2.5-1.5B-apeach")
+    model_name = snapshot_download("Howeee/Qwen2.5-1.5B-apeach", local_files_only=huggingface_hub.constants.HF_HUB_OFFLINE,)
 
     prompts = [
         "Hello, my name is",

tests/e2e/singlecard/pooling/test_embedding.py

@@ -18,6 +18,7 @@
 #
 import pytest
 from modelscope import snapshot_download  # type: ignore[import-untyped]
+import huggingface_hub
 
 from tests.e2e.conftest import HfRunner, VllmRunner
 from tests.e2e.utils import check_embeddings_close
@@ -32,7 +33,7 @@ MODELS = [
 def test_embed_models_correctness(model: str):
     queries = ['What is the capital of China?', 'Explain gravity']
 
-    model_name = snapshot_download(model)
+    model_name = snapshot_download(model, local_files_only=huggingface_hub.constants.HF_HUB_OFFLINE,)
     with VllmRunner(
             model_name,
             runner="pooling",
@@ -60,7 +61,7 @@ def test_embed_models_correctness(model: str):
 def test_bge_m3_correctness():
     queries = ['What is the capital of China?', 'Explain gravity']
 
-    model_name = snapshot_download("BAAI/bge-m3")
+    model_name = snapshot_download("BAAI/bge-m3", local_files_only=huggingface_hub.constants.HF_HUB_OFFLINE,)
     with VllmRunner(
             model_name,
             runner="pooling",

tests/e2e/singlecard/pooling/test_scoring.py

@@ -3,6 +3,7 @@
 import pytest
 import torch
 import torch.nn.functional as F
+import huggingface_hub
 from modelscope import snapshot_download  # type: ignore[import-untyped]
 
 from tests.e2e.conftest import HfRunner, VllmRunner
@@ -31,7 +32,7 @@ DTYPE = "half"
 
 @pytest.fixture(scope="module", params=CROSS_ENCODER_MODELS)
 def model_name(request):
-    yield snapshot_download(request.param)
+    yield snapshot_download(request.param, local_files_only=huggingface_hub.constants.HF_HUB_OFFLINE,)
 
 def test_cross_encoder_score_1_to_1(model_name):
     text_pair = [TEXTS_1[0], TEXTS_2[0]]
@@ -100,7 +101,7 @@ def test_cross_encoder_score_N_to_N(model_name):
 
 @pytest.fixture(scope="module", params=EMBEDDING_MODELS)
 def emb_model_name(request):
-    yield snapshot_download(request.param)
+    yield snapshot_download(request.param, local_files_only=huggingface_hub.constants.HF_HUB_OFFLINE,)
 
 
 def test_embedding_score_1_to_1(emb_model_name):