#!/usr/bin/env bash

export MXNET_CPU_WORKER_NTHREADS=24

export MXNET_GPU_WORKER_NTHREADS=3

export MXNET_CUDNN_AUTOTUNE_DEFAULT=0

export MXNET_ENGINE_TYPE=ThreadedEnginePerDevice

export MXNET_BACKWARD_DO_MIRROR=1

export MXNET_GPU_MEM_POOL_RESERVE=5

#DATA_DIR=/train/mxnet-train/datasets/faces_emore

DATASET='retina'

#NETWORK=r152

NETWORK=resnest

JOB=arcface-$DATASET

MODELDIR="./models/$NETWORK-$JOB"

mkdir -p "$MODELDIR"

PREFIX="$MODELDIR/model"

LOGFILE="$MODELDIR/log"

#CUDA_VISIBLE_DEVICES='0,1,2,3,4,5,6,7' python -u train_parall.py --dataset=$DATASET --network "$NETWORK" --per-batch-size 42 --lr=0.001 --lr-steps='160000,320000,400000' --pretrained '/train/mxnet-train/face/insightface/recognition/models/resnest-arcface-retina/model2,7' >"$LOGFILE-$DATASET-01" 2>&1 &

#CUDA_VISIBLE_DEVICES='0,1,2,3' python -u train_parall.py --dataset=$DATASET --network "$NETWORK" --per-batch-size 48 --lr=0.01 --lr-steps='160000,320000,400000' >"$LOGFILE-$DATASET-04" 2>&1 &

CUDA_VISIBLE_DEVICES='4,5,6,7' python -u train_parall.py --dataset=$DATASET --network "$NETWORK" --per-batch-size 48 --lr=0.001 --lr-steps='160000,320000,400000' --pretrained '/train/mxnet-train/face/insightface/recognition/models/resnest-arcface-retina/model2,9' >"$LOGFILE-$DATASET-03" 2>&1 &

#CUDA_VISIBLE_DEVICES='2,3' python -u train_parall.py --dataset=$DATASET --network "$NETWORK" --per-batch-size 2 --lr=0.01 --lr-steps='160000,320000,400000' >"$LOGFILE-$DATASET-01" 2>&1 &