forked from PRBonn/ContMAV
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrain.sh
16 lines (14 loc) · 850 Bytes
/
train.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
#!/bin/bash
#SBATCH --job-name=img_seg # Job name
#SBATCH --output=LIACi-04-closed-world_%j.log # Output file
#SBATCH --error=LIACi-04-closed-world_%j.log # Error file
#SBATCH --partition=GPUQ # Partition name for GPU access
#SBATCH --gres=gpu:a100:1 # Request 1 A100 GPU with 40GB
#SBATCH --cpus-per-task=4 # Request 4 CPU cores
#SBATCH --mem=40G # Request 40 GB of RAM
#SBATCH --time=10:00:00 # Set a time limit of 1 hour
# Load any necessary modules or activate your conda environment
module load CUDA/12.1.1
source ~/miniconda3/bin/activate openworld
# Run your Python script
python /cluster/home/abubakb/ContMAV_3/train.py --id "LIACi-03-closed-world-02" --dataset_dir "/cluster/home/abubakb/final_dataset" --num_classes 11 --batch_size 4