From bc3a0c06075050d3de586c543e4ad6a7efc9260e Mon Sep 17 00:00:00 2001 From: Ilya Boytsov Date: Sun, 21 Jun 2020 18:51:21 +0300 Subject: [PATCH] [examples] fixes arguments for summarization finetune scripts (#5157) Authored-by: i.boytsov --- examples/summarization/finetune.sh | 5 +---- examples/summarization/finetune_bart_tiny.sh | 3 +-- examples/summarization/finetune_t5.sh | 1 - 3 files changed, 2 insertions(+), 7 deletions(-) diff --git a/examples/summarization/finetune.sh b/examples/summarization/finetune.sh index ead55c5892..b3ee8c9066 100755 --- a/examples/summarization/finetune.sh +++ b/examples/summarization/finetune.sh @@ -1,7 +1,3 @@ -export OUTPUT_DIR=bart_cnn_finetune - -# Make output directory if it doesn't exist -mkdir -p $OUTPUT_DIR # Add parent directory to python path to access lightning_base.py export PYTHONPATH="../":"${PYTHONPATH}" @@ -9,6 +5,7 @@ export PYTHONPATH="../":"${PYTHONPATH}" # --model_name_or_path=t5-base for t5 +# the proper usage is documented in the README python finetune.py \ --model_name_or_path=facebook/bart-large \ --learning_rate=3e-5 \ diff --git a/examples/summarization/finetune_bart_tiny.sh b/examples/summarization/finetune_bart_tiny.sh index 6799e9bb32..b1277d5d7b 100644 --- a/examples/summarization/finetune_bart_tiny.sh +++ b/examples/summarization/finetune_bart_tiny.sh @@ -16,14 +16,13 @@ mkdir -p $OUTPUT_DIR export PYTHONPATH="../":"${PYTHONPATH}" python finetune.py \ --data_dir=cnn_tiny/ \ ---model_type=bart \ --model_name_or_path=sshleifer/bart-tiny-random \ --learning_rate=3e-5 \ --train_batch_size=2 \ --eval_batch_size=2 \ --output_dir=$OUTPUT_DIR \ --num_train_epochs=1 \ ---n_gpu=0 \ +--gpus=0 \ --do_train $@ rm -rf cnn_tiny diff --git a/examples/summarization/finetune_t5.sh b/examples/summarization/finetune_t5.sh index 6f831fae2b..1a97b08117 100644 --- a/examples/summarization/finetune_t5.sh +++ b/examples/summarization/finetune_t5.sh @@ -11,7 +11,6 @@ export PYTHONPATH="../":"${PYTHONPATH}" python finetune.py \ --data_dir=./cnn-dailymail/cnn_dm \ --model_name_or_path=t5-large \ ---model_type=t5 --learning_rate=3e-5 \ --train_batch_size=4 \ --eval_batch_size=4 \