Clean up existing swapfile before trying to allocate a new one (#6066)

Fix a potential bug from
#6058, here is what I see in
the CI when it happens
https://github.com/pytorch/pytorch/actions/runs/12360769784/job/34497379880?pr=143316#step:5:165

terraform-aws-github-runner/modules/runners-instances/templates/user-data.sh

@@ -138,6 +138,10 @@ fi
 
 ${post_install}
 
+# Cleanup any existing swapfile just to be sure
+sudo swapoff /swapfile
+sudo rm /swapfile
+# before allocating a new one
 sudo fallocate -l 3G /swapfile
 sudo chmod 600 /swapfile
 sudo mkswap /swapfile