run.sh

#!/usr/bin/env bash
set -euo pipefail
export TOKENIZERS_PARALLELISM=False

dry_run=false
if [[ "${1:-}" == "--dry-run" ]]; then
    dry_run=true
fi

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
REPO_ROOT="$SCRIPT_DIR"

# -------- Path to YAML config --------
CONFIG_FILE="${REPO_ROOT}/configs/BRIDGE.yaml"
MODEL_FILE="${REPO_ROOT}/configs/dict_model_path.json"
MODEL_FILE_EXAMPLE="${REPO_ROOT}/configs/dict_model_path.example.json"

# -------- Model Name --------
model_name=Llama-3.1-8B-Instruct

# -------- GPU VISIBILITY --------
gpus=0,1,2,3
export CUDA_VISIBLE_DEVICES=$gpus

# -------- Log output --------
dir_log="${REPO_ROOT}/log"
mkdir -p "$dir_log"

# -------- Sanity checks --------
if [[ ! -f "$CONFIG_FILE" ]]; then
    echo "Config file not found: $CONFIG_FILE"
    exit 1
fi
if [[ ! -f "$MODEL_FILE" && "$dry_run" == true && -f "$MODEL_FILE_EXAMPLE" ]]; then
    MODEL_FILE="$MODEL_FILE_EXAMPLE"
elif [[ ! -f "$MODEL_FILE" ]]; then
    echo "Model mapping file not found: $MODEL_FILE"
    echo "Create it from configs/dict_model_path.example.json before running inference."
    exit 1
fi

# -------- Run --------
now=$(date +"%m-%d_%H-%M-%S")
log_file="$dir_log/${model_name}.${now}.log"
cd "$REPO_ROOT"

cmd=(
    python main.py
    --model_name "$model_name"
    --gpus "$gpus"
    --config "$CONFIG_FILE"
    --model_file "$MODEL_FILE"
)

if [[ "$dry_run" == true ]]; then
    echo "Dry run. Command:"
    printf ' %q' "${cmd[@]}"
    echo
    echo "Log file: $log_file"
    exit 0
fi

nohup "${cmd[@]}" > "$log_file" 2>&1 &

# -------- Log output --------
echo "Job started, log in $log_file"