Create hf_trainer.sh

examples/transformers/hf_trainer.sh

@@ -0,0 +1,15 @@
+#!/bin/bash
+#SBATCH --job-name=hf_trainer                   # create a short name for your job
+#SBATCH --output="hf_trainer-%j.out" # %j will be replaced by the slurm jobID
+#SBATCH --nodes=1                         # node count
+#SBATCH --ntasks=1                        # total number of tasks across all nodes
+#SBATCH --cpus-per-task=4                 # cpu-cores per task (>1 if multi-threaded tasks)
+#SBATCH --gres=gpu:nvidia-rtx-a6000:1                      # number of gpus per node
+#SBATCH --mem-per-cpu=2G
+source virtual-venv/bin/activate
+
+echo "User can use the local tmp dir $TMPDIR"
+
+python hf_classification_trainer.py 
+
+deactivate