ENH: xinference python 3.13 support

.github/workflows/python.yaml

@@ -73,14 +73,21 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ "ubuntu-latest", "macos-13", "windows-latest" ]
-        python-version: [ "3.9", "3.10", "3.11", "3.12" ]
+        os: [ "ubuntu-latest", "macos-13", "windows-latest", "macos-14" ]
+        python-version: [ "3.9", "3.10", "3.11", "3.12", "3.13" ]
         module: [ "xinference" ]
         exclude:
           - { os: macos-13, python-version: 3.10 }
           - { os: macos-13, python-version: 3.11 }
+          - { os: macos-13, python-version: 3.12 }
+          - { os: macos-13, python-version: 3.13 }
+          - { os: macos-14, python-version: 3.9 }
+          - { os: macos-14, python-version: 3.10 }
+          - { os: macos-14, python-version: 3.11 }
+          - { os: macos-14, python-version: 3.12 }
           - { os: windows-latest, python-version: 3.10 }
           - { os: windows-latest, python-version: 3.11 }
+          - { os: windows-latest, python-version: 3.12 }
         include:
           - { os: self-hosted, module: gpu, python-version: 3.9}
           - { os: macos-latest, module: metal, python-version: "3.10" }
@@ -99,15 +106,21 @@ jobs:
           python-version: ${{ matrix.python-version }}
           activate-environment: ${{ env.CONDA_ENV }}
 
-      # Important for python == 3.12
+      # Important for python == 3.12 and 3.13
       - name: Update pip and setuptools
-        if: ${{ matrix.python-version == '3.12' }}
+        if: ${{ matrix.python-version == '3.12' || matrix.python-version == '3.13' }}
         run: |
           python -m pip install -U pip setuptools
 
+      # Install torch for Python 3.13 using nightly builds
+      - name: Install torch for Python 3.13
+        if: ${{ matrix.python-version == '3.13'}}
+        run: |
+          python -m pip install torch torchvision torchaudio
+
       - name: Install numpy
         if: |
-          (startsWith(matrix.os, 'macos') && (matrix.python-version == '3.12' || matrix.python-version == '3.9')) || 
+          (startsWith(matrix.os, 'macos') && (matrix.python-version == '3.13' || matrix.python-version == '3.9')) || 
           (startsWith(matrix.os, 'windows') && matrix.python-version == '3.9')
         run: |
           python -m pip install "numpy<2"
@@ -139,7 +152,9 @@ jobs:
             pip install "transformers<4.49"
             pip install attrdict
             pip install "timm>=0.9.16"
-            pip install torch torchvision
+            if [ "${{ matrix.python-version }}" != "3.13" ]; then
+              pip install torch torchvision
+            fi
             pip install accelerate
             pip install sentencepiece
             pip install transformers_stream_generator
@@ -158,9 +173,21 @@ jobs:
           fi
         working-directory: .
 
+      - name: Clean up disk
+        if: |
+          (startsWith(matrix.os, 'ubuntu'))
+        run: |
+          sudo rm -rf /usr/share/dotnet
+          sudo rm -rf /usr/local/lib/android
+          sudo rm -rf /opt/ghc
+          sudo apt-get clean
+          sudo rm -rf /var/lib/apt/lists/*
+          df -h 
+
       - name: Test with pytest
         env:
           MODULE: ${{ matrix.module }}
+          PYTORCH_MPS_HIGH_WATERMARK_RATIO: 1.0
         run: |
           if [ "$MODULE" == "gpu" ]; then
             ${{ env.SELF_HOST_PYTHON }} -m pip install -U -e ".[audio]"

xinference/core/tests/test_metrics.py

@@ -124,6 +124,7 @@ async def test_disable_metrics_exporter_server(disable_metrics, setup_cluster):
         requests.get(metrics_exporter_address)
 
 
+@pytest.mark.timeout(300)  # 5 minutes timeout to prevent hanging in Python 3.13
 async def test_metrics_exporter_data(setup_cluster):
     endpoint, metrics_exporter_address, supervisor_address = setup_cluster
 

xinference/model/embedding/tests/test_embedding_models.py

@@ -249,9 +249,27 @@ def test_register_fault_embedding():
 
     with pytest.warns(UserWarning) as record:
         _install()
-    assert any(
-        "Invalid model URI /new_data/cache/gte-Qwen2" in str(r.message) for r in record
-    )
+
+    # Check for warning message containing the invalid model URI error
+    # The warning format is: "{user_defined_embedding_dir}/{f} has error, {e}"
+    # where e contains the ValueError message
+    found_warning = False
+    for warning in record:
+        message = str(warning.message)
+        if (
+            "has error" in message
+            and (
+                "Invalid model URI" in message
+                or "Model URI cannot be a relative path" in message
+            )
+            and "/new_data/cache/gte-Qwen2" in message
+        ):
+            found_warning = True
+            break
+
+    assert (
+        found_warning
+    ), f"Expected warning about invalid model URI not found. Warnings: {[str(w.message) for w in record]}"
 
 
 def test_convert_ids_to_tokens():