#!/bin/bash
SHELL_FOLDER=$(dirname $(readlink -f "$0"))
echo "cd to "$SHELL_FOLDER

cd $SHELL_FOLDER
######################################################
# export NCCL_DEBUG=INFO
# export NCCL_DEBUG_SUBSYS=ALL

# export NCCL_SOCKET_IFNAME=eth0
export NCCL_IB_DISABLE=1
# export OMP_NUM_THREADS=1
######################################################

export LOCAL_RANK=0

# export WORLD_SIZE=2
# export MASTER_ADDR='1.0.0.0'
# export MASTER_PORT='12345'
# export RANK=0

MIN_SIZE=$WORLD_SIZE
MAX_SIZE=$WORLD_SIZE
TRAINERS_PER_NODE=1
JOB_ID=ddppod

master_addr=$MASTER_ADDR
port=$MASTER_PORT
HOST_NODE_ADDR=${master_addr}":"${port}
RANK=$RANK

echo "current settings"
echo "WORLD_SIZE: "$WORLD_SIZE
echo "MASTER_ADDR: "$MASTER_ADDR
echo "MASTER_PORT: "$MASTER_PORT
echo "RANK: "$RANK
echo "LOCAL_RANK: "$LOCAL_RANK
echo "TEST: "$TEST
echo "job_data_root": $JOB_DATA_ROOT

echo "running python"

python TorchNoduleSeg/test.py \
        --master_ip ${master_addr} \
        --port $port \
        --world_size $WORLD_SIZE \
        --rank $RANK \
        --job_data_root $JOB_DATA_ROOT