virtualenv text-mosaic-augs -p python3 pip install git+https://github.com/huggingface/transformers pip install nltk datasets pip install rouge_score
python run_summarization.py --model_name_or_path t5-small --do_train --do_eval --dataset_name xsum --source_prefix "summarize: " --output_dir "/cs/labs/gabis/ednussi/thesis_small_tasks/sumri_res_test" --per_device_train_batch_size=4 --per_device_eval_batch_size=4 --overwrite_output_dir --predict_with_generate --warmup_ratio=0.1 --max_train_samples 1000000 --num_train_epochs=10 --seed 42 --aug crop
python run_ner.py --model_name_or_path bert-base-uncased --dataset_name wnut_17 --do_train --do_eval --warmup_ratio=0.1 --overwrite_output_dir --num_train_epochs=10 --per_gpu_train_batch_size=8 --seed 42 --max_train_samples 64 --output_dir test_res --aug concat
python run_qa.py --model_name_or_path roberta-base --do_train --do_eval --dataset_name $d --output_dir $OUTPUTDIR --per_device_train_batch_size=4 --per_device_eval_batch_size=4 --overwrite_output_dir --warmup_ratio=0.1 --max_train_samples $i --num_train_epochs=10 --seed $j --aug $aug --save_steps=50000
| Dataset | Full Dataset |
|---|---|
| wnut_17 | 3394 |
| conll2003 | 14041 |
| ncbi_disease | 5433 |
| species_800 | 5734 |
| bc2gm_corpus | 12501 |
| Dataset | Full Dataset |
|---|---|
| xsum | 204045 |
| samsum | 14732 |
| Dataset | Full Dataset |
|---|---|
| squad | 87599 |
| hotpot_qa | 90447 |