|
#!/bin/bash |
|
|
|
|
|
export NCCL_P2P_LEVEL=NVL |
|
echo "dataset $1, model dir $2, input type $3, describe $4, lr $5, lr_LXM $6, batch size $7, wiki num $8, gpu_num $9 " |
|
|
|
export dataset=$1 |
|
export model_dir=$2 |
|
mkdir $model_dir |
|
export input_type=$3 |
|
|
|
export describe=$4 |
|
export lr=$5 |
|
export lr_LXM=$6 |
|
export batch_size=$7 |
|
|
|
export wiki_num=$8 |
|
export gpu_num=$9 |
|
ports=(`echo $METIS_WORKER_0_PORT | tr ',' ' '`) |
|
port=${ports[0]} |
|
|
|
echo "total workers: ${ARNOLD_WORKER_NUM}" |
|
echo "cur worker id: ${ARNOLD_ID}" |
|
echo "gpus per worker: ${ARNOLD_WORKER_GPU}" |
|
echo "master ip: ${METIS_WORKER_0_HOST}" |
|
echo "master port: ${port}" |
|
|
|
|
|
|
|
export OMP_NUM_THREADS=8 |
|
export NCCL_IB_DISABLE=0 |
|
export NCCL_IB_GID_INDEX=3 |
|
export NCCL_IB_HCA=${ARNOLD_RDMA_DEVICE} |
|
export NCCL_SOCKET_IFNAME=eth0 |
|
|
|
python3 -m torch.distributed.launch --nproc_per_node $gpu_num \ |
|
--nnodes=${ARNOLD_WORKER_NUM} --node_rank=${ARNOLD_ID} --master_addr=${METIS_WORKER_0_HOST} --master_port ${port} \ |
|
train4LXMT5_DDP.py \ |
|
--dataset $dataset \ |
|
--model_dir $model_dir \ |
|
--input_type $input_type \ |
|
--describe $describe \ |
|
--learning_rate $lr \ |
|
--learning_rate_LXM $lr_LXM \ |
|
--validate \ |
|
--batch_size $batch_size \ |
|
--num_wiki $wiki_num \ |
|
--pretrain |
|
|