datasets/LogiQA2.0/logiqa2nli/scripts/multirun.sh

export DATA_DIR=../DATA
export TASK_NAME=QA2NLI
python ../run_nli.py  --model_type bert  --model_name_or_path bert-base-uncased --task_name $TASK_NAME --do_train --eval_all_checkpoints --do_eval --do_lower_case --data_dir $DATA_DIR/$TASK_NAME --max_seq_length 128 --per_gpu_eval_batch_size=16 --per_gpu_train_batch_size=8   --gradient_accumulation_steps 3 --logging_steps 5000 --save_steps 5000 --eval_all_checkpoints --learning_rate 2e-5   --num_train_epochs 2.0 --output_dir ./tmp/$TASK_NAME/bert-base/
#python ../run_nli.py  --model_type bert  --model_name_or_path bert-large-uncased --task_name $TASK_NAME --do_train --evaluate_during_training --do_eval --do_lower_case --data_dir $DATA_DIR/$TASK_NAME --max_seq_length 128 --per_gpu_eval_batch_size=16 --per_gpu_train_batch_size=8   --gradient_accumulation_steps 1 --learning_rate 2e-5 --save_steps 200 --adam_epsilon 1e-6 --no_clip_grad_norm --warmup_proportion 0.1 --num_train_epochs 5.0 --output_dir ./tmp/$TASK_NAME/bertlarge/
python ../run_nli.py  --model_type roberta  --model_name_or_path roberta-base --task_name $TASK_NAME  --do_train --do_eval --eval_all_checkpoints --do_lower_case --data_dir $DATA_DIR/$TASK_NAME --max_seq_length 256 --per_gpu_eval_batch_size=16 --per_gpu_train_batch_size=8   --gradient_accumulation_steps 3 --logging_steps 5000 --save_steps 5000 --eval_all_checkpoints --learning_rate 1e-5  --num_train_epochs 2.0 --output_dir ./tmp/$TASK_NAME/roberta/
#python ../run_nli.py  --model_type roberta  --model_name_or_path /home/bimu/PycharmProjects/liu_nli/tmp-1/QNLI/roberta/ --task_name $TASK_NAME --do_train  --do_eval --do_lower_case --data_dir $DATA_DIR/$TASK_NAME --max_seq_length 128 --per_gpu_eval_batch_size=16 --per_gpu_train_batch_size=16   --gradient_accumulation_steps 2 --learning_rate 2e-5  --num_train_epochs 5.0 --output_dir ./tmp/$TASK_NAME/roberta/
#python ../run_nli.py  --model_type xlnet  --model_name_or_path xlnet-base-cased --task_name $TASK_NAME --do_train --do_eval --eval_all_checkpoints --do_lower_case --data_dir $DATA_DIR/$TASK_NAME --max_seq_length 128 --per_gpu_eval_batch_size=16 --per_gpu_train_batch_size=8   --gradient_accumulation_steps 3 --logging_steps 500 --save_steps 500 --eval_all_checkpoints --learning_rate 2e-5  --adam_epsilon 1e-6 --num_train_epochs 5.0 --output_dir ./tmp/$TASK_NAME/xlnet/
#python ../run_nli.py  --model_type bert  --model_name_or_path bert-base-uncased --task_name $TASK_NAME  --do_train --do_eval --do_lower_case --data_dir $DATA_DIR/$TASK_NAME --max_seq_length 128 --per_gpu_eval_batch_size=16 --per_gpu_train_batch_size=16   --gradient_accumulation_steps 2 --logging_steps 500 --save_steps 500 --eval_all_checkpoints --learning_rate 2e-5   --num_train_epochs 5.0 --output_dir ./tmp/$TASK_NAME/bert-base/