first commit
This commit is contained in:
+22
@@ -0,0 +1,22 @@
|
||||
#!/bin/bash
|
||||
#SBATCH --job-name=byt5_nli
|
||||
#SBATCH --output=logs/train_%j.out
|
||||
#SBATCH --error=logs/train_%j.err
|
||||
#SBATCH --gres=gpu:1
|
||||
#SBATCH --cpus-per-task=4
|
||||
#SBATCH --mem=17000M
|
||||
#SBATCH --partition=samsung
|
||||
|
||||
# Required: Activate your environment using the HPC-specific miniconda path
|
||||
source /opt/miniconda3/etc/profile.d/conda.sh
|
||||
conda activate /home/kent_joseph_palima/kent-env-cu11
|
||||
|
||||
# Force Hugging Face libraries to stay offline
|
||||
export TRANSFORMERS_OFFLINE=1
|
||||
export HF_DATASETS_OFFLINE=1
|
||||
|
||||
# Execute the training script using srun as per HPC guidelines
|
||||
srun python train_modified.py
|
||||
|
||||
# Optional: Check GPU status after training
|
||||
# srun nvidia-smi
|
||||
Reference in New Issue
Block a user