125m model

jobscript.sh

 #!/bin/bash
 #SBATCH --account=opengptx-elm
 #SBATCH --partition=booster
-#SBATCH --job-name=opt175b
-#SBATCH --nodes=48
+#SBATCH --job-name=opt125m_test
+#SBATCH --nodes=4
 #SBATCH --hint=nomultithread         
 #SBATCH --ntasks-per-node=4          
 #SBATCH --cpus-per-task=12           # number of cores per tasks
 #SBATCH --gres=gpu:4                 # number of gpus
-#SBATCH --time=00:15:00              # maximum execution time (HH:MM:SS)
+#SBATCH --time=00:10:00              # maximum execution time (HH:MM:SS)
 #SBATCH --output=slurmLog/%x-%j.out  # output file name
 #SBATCH --error=slurmLog/%x-%j.err   # error file name
 
@@ -25,7 +25,7 @@ opt-baselines -n "$SLURM_NNODES" -g 4  \
     --account opengptx-elm \
     --partition booster \
     --prefix "$SLURM_JOB_NAME" \
-    --model-size 175b \
+    --model-size 125m \
     --juwelsbooster \
     --data "$DATA_PATH" \
     --ntasks-per-node 4 \
@@ -34,7 +34,7 @@ opt-baselines -n "$SLURM_NNODES" -g 4  \
     --tensorboard-logdir "$TENSORBOARD_PATH" \
     --no-save-dir \
     --snapshot-root "$ROOT_OUTPUT_DIR" \
-    --time 15  \
+    --time 10  \
     --no-wandb \
     --cpu-bind socket \
     --salloc