folktexts/JZay_exec.sh at main · SebastienMelo/folktexts · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
#!/bin/bash

#SBATCH --job-name=folktexts-gpu
#SBATCH -o outslurm/gpu/job%A_%a.out
#SBATCH -e outslurm/gpu/job%A_%a.err


#SBATCH --gres=gpu:1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=24

#SBATCH -A nuj@h100
#SBATCH -C h100
#SBATCH --hint=nomultithread
#SBATCH --time=90:00:00
#SBATCH --qos=qos_gpu_h100-t4

#SBATCH --array=0-15

echo "------------------------------------------------"
echo "Slurm Job ID: $SLURM_JOB_ID"
echo "Run on host: "`hostname`
echo "Operating system: "`uname -s`
echo "Username: "`whoami`
echo "Started at: "`date`
echo "------------------------------------------------"


module purge
module load arch/h100
module load pytorch-gpu/py3/2.7.0


model_dir=$DSDIR/HuggingFace_Models/

A_VALUES=('meta-llama/Meta-Llama-3-70B' 'meta-llama/Meta-Llama-3-70B-Instruct' 'meta-llama/Meta-Llama-3-8B' 'meta-llama/Meta-Llama-3-8B-Instruct')

T_VALUES=('ACSIncome' 'ACSPublicCoverage' 'ACSMobility' 'ACSTravelTime')


NUM_MODELS=${#A_VALUES[@]}
NUM_TASKS=${#T_VALUES[@]}

MODEL_IDX=$((SLURM_ARRAY_TASK_ID / NUM_TASKS))
TASK_IDX=$((SLURM_ARRAY_TASK_ID % NUM_TASKS))

python -m folktexts.cli.run_acs_benchmark --model "${model_dir}${A_VALUES[$MODEL_IDX]}" --task "${T_VALUES[$TASK_IDX]}" --data-dir data --results-dir folktexts-results-ICML --batch-size 16
# srun run_acs_benchmark --model ${A_VALUES[$SLURM_ARRAY_TASK_ID]} --task ACSIncome --results-dir folktexts-results --data-dir data