[CI/UT] Unify model usage via ModelScope in CI (#1207)

### What this PR does / why we need it?
Unify Model Usage via ModelScope

### Does this PR introduce _any_ user-facing change?
No

### How was this patch tested?
CI passed

Signed-off-by: hfadzxy <starmoon_zhang@163.com>

.github/workflows/accuracy_test.yaml

@@ -125,10 +125,9 @@ jobs:
     container:
       image: m.daocloud.io/quay.io/ascend/cann:8.1.rc1-910b-ubuntu22.04-py3.10
       env:
-        HF_ENDPOINT: https://hf-mirror.com
-        HF_TOKEN: ${{ secrets.HF_TOKEN }}
         DATASET_SOURCE: ModelScope
         VLLM_USE_MODELSCOPE: True
+        USE_MODELSCOPE_HUB: 1
         # 1. If version specified (work_dispatch), do specified branch accuracy test
         # 2. If no version (labeled PR), do accuracy test by default ref:
         # The branch, tag or SHA to checkout. When checking out the repository that

.github/workflows/nightly_benchmarks.yaml

@@ -69,8 +69,7 @@ jobs:
         --device /dev/devmm_svm
         --device /dev/hisi_hdc
       env:
-        HF_ENDPOINT: https://hf-mirror.com
+        VLLM_USE_MODELSCOPE: True
-        HF_TOKEN: ${{ secrets.HF_TOKEN }}
         ES_OM_DOMAIN: ${{ secrets.ES_OM_DOMAIN }}
         ES_OM_AUTHORIZATION: ${{ secrets.ES_OM_AUTHORIZATION }}
         VLLM_USE_V1: ${{ matrix.vllm_use_v1 }}

.github/workflows/vllm_ascend_test.yaml

@@ -209,6 +209,7 @@ jobs:
       image: m.daocloud.io/quay.io/ascend/cann:8.1.rc1-910b-ubuntu22.04-py3.10
       env:
         VLLM_LOGGING_LEVEL: ERROR
+        VLLM_USE_MODELSCOPE: True
     steps:
       - name: Check npu and CANN info
         run: |
@@ -257,9 +258,7 @@ jobs:
           VLLM_USE_MODELSCOPE: True
         run: |
           pytest -sv tests/e2e/singlecard/test_offline_inference.py
-          # TODO: switch hf to modelscope
+          pytest -sv tests/e2e/singlecard/test_ilama_lora.py
-          VLLM_USE_MODELSCOPE=False HF_ENDPOINT=https://hf-mirror.com \
-            pytest -sv tests/e2e/singlecard/test_ilama_lora.py
           pytest -sv tests/e2e/singlecard/test_guided_decoding.py
           pytest -sv tests/e2e/singlecard/test_camem.py
           pytest -sv tests/e2e/singlecard/test_embedding.py
@@ -277,9 +276,7 @@ jobs:
           VLLM_USE_MODELSCOPE: True
         run: |
           pytest -sv tests/e2e/singlecard/test_offline_inference.py
-          # TODO: switch hf to modelscope
+          pytest -sv tests/e2e/singlecard/test_ilama_lora.py
-          VLLM_USE_MODELSCOPE=False HF_ENDPOINT=https://hf-mirror.com \
-            pytest -sv tests/e2e/singlecard/test_ilama_lora.py
           pytest -sv tests/e2e/singlecard/test_guided_decoding.py
           pytest -sv tests/e2e/singlecard/test_camem.py
           pytest -sv tests/e2e/singlecard/test_prompt_embedding.py
@@ -357,9 +354,7 @@ jobs:
           VLLM_WORKER_MULTIPROC_METHOD: spawn
           VLLM_USE_MODELSCOPE: True
         run: |
-          # TODO: switch hf to modelscope
+          pytest -sv tests/e2e/multicard/test_ilama_lora_tp2.py
-          VLLM_USE_MODELSCOPE=False HF_ENDPOINT=https://hf-mirror.com \
-            pytest -sv tests/e2e/multicard/test_ilama_lora_tp2.py
           # Fixme: run VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py will raise error.
           # To avoid oom, we need to run the test in a single process.
           pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek_multistream_moe
@@ -380,9 +375,7 @@ jobs:
           VLLM_USE_V1: 0
           VLLM_USE_MODELSCOPE: True
         run: |
-          # TODO: switch hf to modelscope
+          pytest -sv tests/e2e/multicard/test_ilama_lora_tp2.py
-          VLLM_USE_MODELSCOPE=False HF_ENDPOINT=https://hf-mirror.com \
-            pytest -sv tests/e2e/multicard/test_ilama_lora_tp2.py
           # Fixme: run VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py will raise error.
           # To avoid oom, we need to run the test in a single process.
           pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_QwQ

.github/workflows/vllm_ascend_test_long_term.yaml

@@ -50,9 +50,8 @@ jobs:
       # TODO(yikun): Remove m.daocloud.io prefix when infra proxy ready
       image: m.daocloud.io/quay.io/ascend/cann:8.1.rc1-910b-ubuntu22.04-py3.10
       env:
-        HF_ENDPOINT: https://hf-mirror.com
-        HF_TOKEN: ${{ secrets.HF_TOKEN }}
         VLLM_LOGGING_LEVEL: ERROR
+        VLLM_USE_MODELSCOPE: True
     steps:
       - name: Check npu and CANN info
         run: |

.github/workflows/vllm_ascend_test_pd.yaml

@@ -64,8 +64,7 @@ jobs:
         --device /dev/devmm_svm
         --device /dev/hisi_hdc
       env:
-        HF_ENDPOINT: https://hf-mirror.com
+        VLLM_USE_MODELSCOPE: True
-        HF_TOKEN: ${{ secrets.HF_TOKEN }}
     steps:
       - name: Check npu and CANN info
         run: |

benchmarks/scripts/run-performance-benchmarks.sh

@@ -295,7 +295,7 @@ main() {
   export VLLM_LOG_LEVEL="WARNING"
   
   # set env
-  export HF_ENDPOINT="https://hf-mirror.com"
+  export VLLM_USE_MODELSCOPE=True
 
   # prepare for benchmarking
   cd benchmarks || exit 1

tests/conftest.py

@@ -25,7 +25,7 @@ from typing import Any, List, Optional, Tuple, TypeVar, Union
 import numpy as np
 import pytest
 import torch
-from huggingface_hub import snapshot_download
+from modelscope import snapshot_download  # type: ignore[import-untyped]
 from PIL import Image
 from torch import nn
 from transformers import (AutoConfig, AutoModelForCausalLM, AutoTokenizer,
@@ -387,7 +387,7 @@ def example_prompts() -> list[str]:
 
 @pytest.fixture(scope="session")
 def ilama_lora_files():
-    return snapshot_download(repo_id="jeeejeee/ilama-text2sql-spider")
+    return snapshot_download(repo_id="vllm-ascend/ilama-text2sql-spider")
 
 
 class HfRunner:

tests/e2e/multicard/test_ilama_lora_tp2.py

@@ -1,4 +1,5 @@
 import pytest
+from modelscope import snapshot_download  # type: ignore
 
 from tests.conftest import VllmRunner
 from tests.e2e.singlecard.test_ilama_lora import (EXPECTED_LORA_OUTPUT,
@@ -7,7 +8,7 @@ from tests.e2e.singlecard.test_ilama_lora import (EXPECTED_LORA_OUTPUT,
 
 @pytest.mark.parametrize("distributed_executor_backend", ["mp"])
 def test_ilama_lora_tp2(distributed_executor_backend, ilama_lora_files):
-    with VllmRunner(model_name=MODEL_PATH,
+    with VllmRunner(snapshot_download(MODEL_PATH),
                     enable_lora=True,
                     max_loras=4,
                     max_model_len=1024,

tests/e2e/singlecard/test_ilama_lora.py

@@ -1,10 +1,11 @@
 # SPDX-License-Identifier: Apache-2.0
 import vllm
+from modelscope import snapshot_download  # type: ignore
 from vllm.lora.request import LoRARequest
 
 from tests.conftest import VllmRunner
 
-MODEL_PATH = "ArthurZ/ilama-3.2-1B"
+MODEL_PATH = "vllm-ascend/ilama-3.2-1B"
 
 PROMPT_TEMPLATE = """I want you to act as a SQL terminal in front of an example database, you need only to return the sql command to me.Below is an instruction that describes a task, Write a response that appropriately completes the request.\n"\n##Instruction:\nconcert_singer contains tables such as stadium, singer, concert, singer_in_concert. Table stadium has columns such as Stadium_ID, Location, Name, Capacity, Highest, Lowest, Average. Stadium_ID is the primary key.\nTable singer has columns such as Singer_ID, Name, Country, Song_Name, Song_release_year, Age, Is_male. Singer_ID is the primary key.\nTable concert has columns such as concert_ID, concert_Name, Theme, Stadium_ID, Year. concert_ID is the primary key.\nTable singer_in_concert has columns such as concert_ID, Singer_ID. concert_ID is the primary key.\nThe Stadium_ID of concert is the foreign key of Stadium_ID of stadium.\nThe Singer_ID of singer_in_concert is the foreign key of Singer_ID of singer.\nThe concert_ID of singer_in_concert is the foreign key of concert_ID of concert.\n\n###Input:\n{query}\n\n###Response:"""  # noqa: E501
 
@@ -44,7 +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(model_name=MODEL_PATH,
+    with VllmRunner(snapshot_download(MODEL_PATH),
                     enable_lora=True,
                     max_loras=4,
                     max_model_len=1024,