ci: fix rate-limit of huggingface with hf auth login (#10947)

.github/workflows/pr-test.yml

@@ -103,6 +103,8 @@ jobs:
     needs: [check-changes, sgl-kernel-build-wheels]
     if: needs.check-changes.outputs.sgl_kernel == 'true'
     runs-on: 1-gpu-runner
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
     steps:
       - uses: actions/checkout@v4
 
@@ -191,6 +193,8 @@ jobs:
     if: always() && !failure() && !cancelled() &&
         ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
     runs-on: 1-gpu-runner
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
     strategy:
       fail-fast: false
       matrix:
@@ -222,6 +226,8 @@ jobs:
     if: always() && !failure() && !cancelled() &&
         ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
     runs-on: 2-gpu-runner
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
     strategy:
       fail-fast: false
       matrix:
@@ -253,6 +259,8 @@ jobs:
     if: always() && !failure() && !cancelled() &&
         ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
     runs-on: 4-gpu-runner
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
     strategy:
       fail-fast: false
       matrix:
@@ -284,6 +292,8 @@ jobs:
     if: always() && !failure() && !cancelled() &&
         ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
     runs-on: 8-gpu-runner
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
     strategy:
       fail-fast: false
       matrix:
@@ -315,6 +325,8 @@ jobs:
     if: always() && !failure() && !cancelled() &&
         ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
     runs-on: 1-gpu-runner
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
@@ -374,6 +386,8 @@ jobs:
     if: always() && !failure() && !cancelled() &&
         ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
     runs-on: 1-gpu-runner
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
@@ -425,6 +439,8 @@ jobs:
     if: always() && !failure() && !cancelled() &&
         ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
     runs-on: 2-gpu-runner
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
@@ -482,6 +498,8 @@ jobs:
     if: always() && !failure() && !cancelled() &&
         ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
     runs-on: 1-gpu-runner
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
@@ -512,6 +530,8 @@ jobs:
     if: always() && !failure() && !cancelled() &&
         ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
     runs-on: 2-gpu-runner
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
@@ -542,6 +562,8 @@ jobs:
     if: always() && !failure() && !cancelled() &&
         ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
     runs-on: 4-gpu-runner
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
@@ -569,6 +591,8 @@ jobs:
     if: always() && !failure() && !cancelled() &&
         ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
     runs-on: 8-gpu-runner
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
@@ -596,6 +620,8 @@ jobs:
     if: always() && !failure() && !cancelled() &&
         ((needs.check-changes.outputs.main_package == 'true') || (needs.check-changes.outputs.sgl_kernel == 'true'))
     runs-on: 4-b200-runner
+    env:
+      HF_TOKEN: ${{ secrets.HF_TOKEN }}
     strategy:
       fail-fast: false
     steps:

scripts/ci/ci_install_dependency.sh

@@ -91,3 +91,9 @@ fi
 
 # Show current packages
 $PIP_CMD list
+
+
+if [ -n "${HF_TOKEN:-}" ]; then
+    $PIP_CMD install -U "huggingface_hub[cli]" $PIP_INSTALL_SUFFIX
+    hf auth login --token $HF_TOKEN
+fi

test/srt/run_suite.py

@@ -35,7 +35,7 @@ suites = {
         # TestFile("models/test_grok_models.py", 60),  # Disabled due to illegal memory access
         TestFile("models/test_qwen_models.py", 82),
         TestFile("models/test_reward_models.py", 132),
-        TestFile("models/test_vlm_models.py", 437),
+        TestFile("models/test_vlm_models.py", 741),
         TestFile("models/test_transformers_models.py", 320),
         TestFile("openai_server/basic/test_protocol.py", 10),
         TestFile("openai_server/basic/test_serving_chat.py", 10),
@@ -93,7 +93,7 @@ suites = {
         TestFile("test_multi_tokenizer.py", 230),
         TestFile("test_no_chunked_prefill.py", 108),
         TestFile("test_no_overlap_scheduler.py", 234),
-        TestFile("test_original_logprobs.py", 200),
+        TestFile("test_original_logprobs.py", 41),
         TestFile("test_penalty.py", 41),
         TestFile("test_page_size.py", 60),
         TestFile("test_priority_scheduling.py", 100),
@@ -244,7 +244,7 @@ suite_amd = {
         TestFile("test_skip_tokenizer_init.py", 117),
         TestFile("test_srt_engine.py", 261),
         TestFile("test_srt_endpoint.py", 130),
-        TestFile("test_torch_compile.py", 76),
+        TestFile("test_torch_compile.py", 169),
         TestFile("test_torch_compile_moe.py", 172),
         TestFile("test_torch_native_attention_backend.py", 123),
         TestFile("test_triton_attention_backend.py", 150),