try nvidia-cuda

ghstack-source-id: af0575dff2fc8db048674e4d59fe09e23421a463
Pull Request resolved: #332

.github/workflows/unit_test_4gpu.yaml

@@ -16,12 +16,20 @@ jobs:
       runner: linux.g5.12xlarge.nvidia.gpu
       gpu-arch-type: cuda
       gpu-arch-version: "12.1"
-      # This image is faster to clone than the default, but it lacks CC needed by triton
-      # (1m25s vs 2m37s)
-      docker-image: "pytorch/pytorch:2.3.0-cuda12.1-cudnn8-runtime"
+      # Trying how much faster the nvidia-cuda image is
+      docker-image: "nvidia/cuda:12.4.1-runtime-ubuntu22.04"
       repository: "pytorch/torchtitan"
       upload-artifact: "outputs"
+        #         ~/miniconda3/bin/conda init bash
       script: |
+        apt update && apt install -y wget
+        mkdir -p ~/miniconda3
+        wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
+        bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
+        rm -rf ~/miniconda3/miniconda.sh
+        source ~/miniconda3/bin/activate
+        conda create -n "test" python=3.10
+        conda activate test
         conda install -y -q git clang clangxx
         export CC=clang
         export CXX=clangxx