Merge branch 'main' into mwittwer/explicit_model_load_parsing

Author: mattwittwer

Dockerfile.sdk

@@ -43,7 +43,7 @@ ARG JAVA_BINDINGS_MAVEN_VERSION=3.8.4
 ARG JAVA_BINDINGS_JAVACPP_PRESETS_TAG=1.5.8
 ARG TRITON_PERF_ANALYZER_BUILD=1
 # DCGM version to install for Model Analyzer
-ARG DCGM_VERSION=3.3.6
+ARG DCGM_VERSION=4.2.3-2
 
 ARG NVIDIA_TRITON_SERVER_SDK_VERSION=unknown
 ARG NVIDIA_BUILD_ID=unknown
@@ -265,7 +265,7 @@ RUN if [ "$TRITON_ENABLE_GPU" = "ON" ]; then \
         curl -o /tmp/cuda-keyring.deb \
         https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/$arch/cuda-keyring_1.1-1_all.deb \
         && apt install /tmp/cuda-keyring.deb && rm /tmp/cuda-keyring.deb && \
-        apt-get update && apt-get install -y datacenter-gpu-manager=1:${DCGM_VERSION}; \
+        apt update && apt install --yes datacenter-gpu-manager-4-dev=1:${DCGM_VERSION}; \
     fi
 
 # Build expects "python" executable (not python3).

build.py

@@ -77,7 +77,7 @@
     "ort_version": "1.22.0",
     "ort_openvino_version": "2025.2.0",
     "standalone_openvino_version": "2025.2.0",
-    "dcgm_version": "3.3.6",
+    "dcgm_version": "4.2.3-2",
     "vllm_version": "0.9.0.1",
     "rhel_py_version": "3.12.3",
 }
@@ -858,7 +858,7 @@ def install_dcgm_libraries(dcgm_version, target_machine):
 # Install DCGM. Steps from https://developer.nvidia.com/dcgm#Downloads
 RUN dnf config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel8/sbsa/cuda-rhel8.repo \\
     && dnf clean expire-cache \\
-    && dnf install -y datacenter-gpu-manager-{}
+    && dnf install --assumeyes datacenter-gpu-manager-4-devel=1:{}
 """.format(
                     dcgm_version, dcgm_version
                 )
@@ -868,7 +868,7 @@ def install_dcgm_libraries(dcgm_version, target_machine):
 # Install DCGM. Steps from https://developer.nvidia.com/dcgm#Downloads
 RUN dnf config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel8/x86_64/cuda-rhel8.repo \\
     && dnf clean expire-cache \\
-    && dnf install -y datacenter-gpu-manager-{}
+    && dnf install --assumeyes datacenter-gpu-manager-4-devel=1:{}
 """.format(
                     dcgm_version, dcgm_version
                 )
@@ -881,8 +881,8 @@ def install_dcgm_libraries(dcgm_version, target_machine):
         https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/sbsa/cuda-keyring_1.1-1_all.deb \\
       && apt install /tmp/cuda-keyring.deb \\
       && rm /tmp/cuda-keyring.deb \\
-      && apt-get update \\
-      && apt-get install -y datacenter-gpu-manager=1:{}
+      && apt update \\
+      && apt install --yes datacenter-gpu-manager-4-dev=1:{}
 """.format(
                     dcgm_version, dcgm_version
                 )
@@ -894,8 +894,8 @@ def install_dcgm_libraries(dcgm_version, target_machine):
           https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/cuda-keyring_1.1-1_all.deb \\
       && apt install /tmp/cuda-keyring.deb \\
       && rm /tmp/cuda-keyring.deb \\
-      && apt-get update \\
-      && apt-get install -y datacenter-gpu-manager=1:{}
+      && apt update \\
+      && apt install --yes datacenter-gpu-manager-4-dev=1:{}
 """.format(
                     dcgm_version, dcgm_version
                 )

compose.py

@@ -298,7 +298,7 @@ def create_argmap(images, skip_pull):
     dcgm_ver = re.search("DCGM_VERSION=([\S]{4,}) ", vars)
     dcgm_version = ""
     if dcgm_ver is None:
-        dcgm_version = "3.3.6"
+        dcgm_version = "4.2.3-2"
         log(
             "WARNING: DCGM version not found from image, installing the earlierst version {}".format(
                 dcgm_version

qa/common/test_util.py

@@ -312,7 +312,7 @@ def check_gpus_compute_capability(min_capability):
 
     import importlib.util
 
-    if importlib.util.find_spec("cuda.core.experimental"):
+    if importlib.util.find_spec("cuda") is not None:
         import cuda.core.experimental as cuda_core_experimental
 
         devices = cuda_core_experimental.system.devices
@@ -325,7 +325,7 @@ def check_gpus_compute_capability(min_capability):
             if compute_capability < min_capability:
                 return False
 
-    elif importlib.util.find_spec("pycuda.driver"):
+    elif importlib.util.find_spec("pycuda") is not None:
         import pycuda.driver as cuda
 
         cuda.init()
@@ -339,6 +339,10 @@ def check_gpus_compute_capability(min_capability):
 
             if compute_capability_value < min_capability:
                 return False
+    else:
+        raise RuntimeError(
+            "No packages found to determine the compute capability. Please check the environment."
+        )
 
     return True