Merge pull request #27 from coreweave/wbrown.cuda13

build: Add disabled CUDA 12.3, add `--allow-downgrades`.

.github/workflows/ubuntu-20.yml

@@ -7,22 +7,6 @@ on:
       - .github/workflows/build.yml
 
 jobs:
-  cu117:
-    uses: ./.github/workflows/build.yml
-    with:
-      folder: .
-      dockerfile: Dockerfile.ubuntu20
-      base-image: nvidia/cuda
-      base-tag: 11.7.1-cudnn8-devel-ubuntu20.04
-      cuda-version-minor: "11.7.1"
-      cuda-version-major: "11.7"
-      nccl-version: 2.14.3-1
-      cuda-samples-version: "11.6"
-      hpcx-version: "2.14"
-      hpcx-nccl-version: "2.16"
-      hpcx-cuda-version: "11"
-      hpcx-mlnx-ofed: "MLNX_OFED_LINUX-5"
-
   cu118:
     uses: ./.github/workflows/build.yml
     with:
@@ -32,7 +16,7 @@ jobs:
       base-tag: 11.8.0-cudnn8-devel-ubuntu20.04
       cuda-version-minor: "11.8.0"
       cuda-version-major: "11.8"
-      nccl-version: 2.16.2-1
+      nccl-version: 2.16.5-1
       cuda-samples-version: "11.6"
       hpcx-version: "2.14"
       hpcx-nccl-version: "2.16"
@@ -48,7 +32,7 @@ jobs:
       base-tag: 12.0.1-cudnn8-devel-ubuntu20.04
       cuda-version-minor: "12.0.1"
       cuda-version-major: "12.0"
-      nccl-version: 2.18.5-1
+      nccl-version: 2.19.3-1
       cuda-samples-version: "12.0"
       hpcx-version: "2.16"
       hpcx-nccl-version: "2.18"
@@ -80,9 +64,25 @@ jobs:
       base-tag: 12.2.2-cudnn8-devel-ubuntu20.04
       cuda-version-minor: "12.2.2"
       cuda-version-major: "12.2"
-      nccl-version: 2.18.5-1
+      nccl-version: 2.19.3-1
       cuda-samples-version: "12.2"
       hpcx-version: "2.16"
       hpcx-nccl-version: "2.18"
       hpcx-cuda-version: "12"
       hpcx-mlnx-ofed: "mlnx_ofed"
+
+#  cu123:
+#    uses: ./.github/workflows/build.yml
+#    with:
+#      folder: .
+#      dockerfile: Dockerfile.ubuntu20
+#      base-image: nvidia/cuda
+#      base-tag: 12.3.0-cudnn8-devel-ubuntu20.04
+#      cuda-version-minor: "12.3.0"
+#      cuda-version-major: "12.3"
+#      nccl-version: 2.19.3-1
+#      cuda-samples-version: "12.3"
+#      hpcx-version: "2.17"
+#      hpcx-nccl-version: "2.19"
+#      hpcx-cuda-version: "12"
+#      hpcx-mlnx-ofed: "mlnx_ofed"

Dockerfile.ubuntu20

@@ -7,15 +7,19 @@ ARG TARGET_NCCL_VERSION=2.14.3-1
 
 ARG DEBIAN_FRONTEND=noninteractive
 RUN apt-get -qq update && \
-        apt-get -qq install -y --allow-change-held-packages --no-install-recommends \
+        apt-get -qq install -y \
+        --allow-change-held-packages \
+        --no-install-recommends \
+        --allow-downgrades \
         build-essential libtool autoconf automake autotools-dev unzip \
         ca-certificates \
         wget curl openssh-server vim environment-modules \
         iputils-ping net-tools \
         libnuma1 libsubunit0 libpci-dev \
         libpmix-dev \
         datacenter-gpu-manager \
-        libnccl2=$TARGET_NCCL_VERSION+cuda${CUDA_VERSION_MAJOR} libnccl-dev=${TARGET_NCCL_VERSION}+cuda${CUDA_VERSION_MAJOR} \
+        libnccl2=$TARGET_NCCL_VERSION+cuda${CUDA_VERSION_MAJOR} \
+        libnccl-dev=${TARGET_NCCL_VERSION}+cuda${CUDA_VERSION_MAJOR} \
         git
 
 # Mellanox OFED (latest)
@@ -63,7 +67,7 @@ ARG HPCX_MLNX_OFED="MLNX_OFED_LINUX-5"
 RUN cd /tmp && \
     export HPCX_DISTRIBUTION="hpcx-v${HPCX_VERSION}-gcc-${HPCX_MLNX_OFED}-ubuntu20.04-cuda${HPCX_CUDA_VERSION}-gdrcopy2-nccl${HPCX_NCCL_VERSION}-x86_64" \
            HPCX_DIR="/opt/hpcx" && \
-    wget -q -O - http://blobstore.s3.ord1.coreweave.com/drivers/${HPCX_DISTRIBUTION}.tbz | tar xjf - && \
+    wget -q -O - http://blobstore.object.ord1.coreweave.com/drivers/${HPCX_DISTRIBUTION}.tbz | tar xjf - && \
     grep -IrlF "/build-result/${HPCX_DISTRIBUTION}" ${HPCX_DISTRIBUTION} | xargs -rd'\n' sed -i -e "s:/build-result/${HPCX_DISTRIBUTION}:${HPCX_DIR}:g" && \
     mv ${HPCX_DISTRIBUTION} ${HPCX_DIR}