jobscript-reranker-run-train

#!/bin/bash

# set a job name
#SBATCH --job-name=reranker
#################

# a file for job output, you can check job progress
#SBATCH --output=reranker_slurm_output_%j.out
#################

# a file for errors
#SBATCH --error=reranker_slurm_output_%j.err
#################

# time needed for job
#SBATCH --time=00:30:00
#################

# gpus per node
#SBATCH --gres=gpu:1
#################

# cpus per job
#SBATCH --cpus-per-task=1
#################

# number of requested nodes
#SBATCH --nodes=1
#################

# memory per node
#SBATCH --mem=8GB
#################

# slurm will send a signal this far out before it kills the job
#SBATCH --signal=USR1@300
#################


# tasks per node
#SBATCH --tasks-per-node=1
#################

module load cuda
#module load tensorflow/2.5.0-py39-cuda112
source venv/bin/activate


srun python3 run_agask.py \
   --save_steps 2000 \
   --max_len 512 \
   --cache_dir cache\
   --per_device_train_batch_size 2 \
   --train_group_size 2 \
   --gradient_accumulation_steps 2 \
   --weight_decay 0.01 \
   --learning_rate 1e-6 \
   --num_train_epochs 10 \
   --dataloader_num_workers 8 \
   --fp16 \
   --do_train \
   --output_dir examples/agask/models/agask_model_custom_params_queries \
   --model_name_or_path /scratch1/koo01a/pt-bert-large-msmarco \
   --train_dir examples/agask/feature_json \
   --overwrite_output_dir