Revert Phoenix test jobs to multi-partition GPU scheduling

sbryngelson · claude · sbryngelson · commit edefc015be0f · 2026-02-14T20:26:51.000-05:00
Keep H200 targeting only for benchmarks; tests should run on any
available GPU partition for faster scheduling.

Co-Authored-By: Claude Opus 4.6 &lt;noreply@anthropic.com&gt;
diff --git a/.github/workflows/phoenix/submit.sh b/.github/workflows/phoenix/submit.sh
@@ -23,8 +23,9 @@ sbatch_cpu_opts="\
 "
 
 sbatch_gpu_opts="\
-#SBATCH --gres=gpu:H200:2
-#SBATCH --ntasks-per-node=8       # Number of cores per node required\
+#SBATCH -p gpu-v100,gpu-a100,gpu-h100,gpu-l40s
+#SBATCH --ntasks-per-node=4       # Number of cores per node required
+#SBATCH -G2\
 "
 
 if [ "$2" = "cpu" ]; then

Original file line number	Diff line number	Diff line change
`@@ -23,8 +23,9 @@ sbatch_cpu_opts="\`
`23`	`23`	`"`
`24`	`24`
`25`	`25`	`sbatch_gpu_opts="\`
`26`		`-#SBATCH --gres=gpu:H200:2`
`27`		`-#SBATCH --ntasks-per-node=8 # Number of cores per node required\`
	`26`	`+#SBATCH -p gpu-v100,gpu-a100,gpu-h100,gpu-l40s`
	`27`	`+#SBATCH --ntasks-per-node=4 # Number of cores per node required`
	`28`	`+#SBATCH -G2\`
`28`	`29`	`"`
`29`	`30`
`30`	`31`	`if [ "$2" = "cpu" ]; then`