export FLAGS_eager_delete_tensor_gb=0.0
export CUDA_VISIBLE_DEVICES=0
CKPT_DIR="./ckpt_sequence_label"
python -u sequence_label.py \
--batch_size 16 \ --weight_decay 0.01 \ --num_epoch 3 \
--checkpoint_dir $CKPT_DIR \
--max_seq_len 128 \
--learning_rate 5e-5 \ --use_pyreader True \ --use_data_parallel True
