run commands within apptainer

abcucberkeley · Oct 28, 2024 · a6cfce2 · a6cfce2
1 parent 0e912f7
commit a6cfce2
Show file tree

Hide file tree

Showing 5 changed files with 21 additions and 23 deletions.
diff --git a/src/ray_check_status.sh b/src/ray_check_status.sh
@@ -1,9 +1,9 @@
 #!/bin/bash
 
 # conda create -n ray -c conda-forge python=3.10.12 "ray-all"=2.37.0 -y
-source ~/.bashrc
-conda activate ray
-echo "Activated conda environment: ray"
+#source ~/.bashrc
+#conda activate ray
+#echo "Activated conda environment: ray"
 
 export LC_ALL=C.UTF-8
 export LANG=C.UTF-8
@@ -27,6 +27,8 @@ while getopts ":e:a:r:" option;do
     esac
 done
 
+echo "Starting apptainer: $env"
+$env bash
 
 TIMEOUT=60 # seconds
 check="ray status --address $cluster_address"

diff --git a/src/ray_lsf_cluster.sh b/src/ray_lsf_cluster.sh
@@ -1,10 +1,5 @@
 #!/bin/bash
 
-# conda create -n ray -c conda-forge python=3.10.12 "ray-all"=2.37.0 -y
-source ~/.bashrc
-conda activate ray
-echo "Activated conda environment: ray"
-
 while getopts ":e:n:c:g:o:w:" option;do
     case "${option}" in
     e) e=${OPTARG}

diff --git a/src/ray_lsf_cluster_parallel.sh b/src/ray_lsf_cluster_parallel.sh
@@ -1,10 +1,5 @@
 #!/bin/bash
 
-# conda create -n ray -c conda-forge python=3.10.12 "ray-all"=2.37.0 -y
-source ~/.bashrc
-conda activate ray
-echo "Activated conda environment: ray"
-
 while getopts ":n:c:g:o:w:" option;do
     case "${option}" in
     n) n=${OPTARG}

diff --git a/src/ray_start_cluster.sh b/src/ray_start_cluster.sh
@@ -1,9 +1,9 @@
 #!/bin/bash
 
 # conda create -n ray -c conda-forge python=3.10.12 "ray-all"=2.37.0 -y
-source ~/.bashrc
-conda activate ray
-echo "Activated conda environment: ray"
+#source ~/.bashrc
+#conda activate ray
+#echo "Activated conda environment: ray"
 
 export LC_ALL=C.UTF-8
 export LANG=C.UTF-8
@@ -41,9 +41,12 @@ done
 
 cluster_address="$ip:$port"
 
+echo "Starting apptainer: $env"
+$env bash
+
 echo "Starting ray head node @ $(hostname) => $cluster_address with CPUs[$cpus] & GPUs [$gpus]"
 job="ray start --head --node-ip-address=$ip --port=$port --dashboard-port=$dashboard_port --dashboard-host=0.0.0.0 --temp-dir=/tmp/ray_symlink --num-cpus=$cpus --num-gpus=$gpus"
-echo $env $job
-$env $job &
+echo $job
+$job &
 
 sleep infinity
diff --git a/src/ray_start_worker.sh b/src/ray_start_worker.sh
@@ -1,9 +1,9 @@
 #!/bin/bash
 
 # conda create -n ray -c conda-forge python=3.10.12 "ray-all"=2.37.0 -y
-source ~/.bashrc
-conda activate ray
-echo "Activated conda environment: ray"
+#source ~/.bashrc
+#conda activate ray
+#echo "Activated conda environment: ray"
 
 export LC_ALL=C.UTF-8
 export LANG=C.UTF-8
@@ -31,10 +31,13 @@ while getopts ":e:a:c:g:" option;do
     esac
 done
 
+echo "Starting apptainer: $env"
+$env bash
+
 echo "Starting ray worker @ $(hostname) with CPUs[$cpus] & GPUs [$gpus] => $cluster_address"
 job="ray start --address=$cluster_address --num-cpus=$cpus --num-gpus=$gpus"
-echo $env $job
-$env $job &
+echo $job
+$job &
 
 echo "Ray worker LSF ID: $LSB_JOBID"
 sleep infinity