From c0dcc5f0f6ce9df0f62828e1a653032ddd50f13f Mon Sep 17 00:00:00 2001
From: yyy-Apple <50064641+yyy-Apple@users.noreply.github.com>
Date: Wed, 2 Mar 2022 00:31:11 +0800
Subject: [PATCH] update README.md

---
 train/README.md                | 2 +-
 train/reproduce/data/README.md | 7 +++++++
 2 files changed, 8 insertions(+), 1 deletion(-)
 create mode 100644 train/reproduce/data/README.md

diff --git a/train/README.md b/train/README.md
index 0b434cb..51b0d5e 100644
--- a/train/README.md
+++ b/train/README.md
@@ -18,7 +18,7 @@ python bart.py --train_file train.json --validation_file val.json --output_dir m
 Then you can use your custom BARTScore for evaluation.
 
 ## Reproduce
-To reproduce our results, please see the folder [`reproduce`](reproduce). Due to limited computing resources, we sharded the BART into multiple GPUs and trained the model. See [`reproduce/finetune.py`](reproduce/finetune.py) for details.
+To reproduce our results, please see the folder [`reproduce`](reproduce). Due to limited computing resources, we sharded the BART into multiple GPUs and trained the model, please see [`reproduce/finetune.py`](reproduce/finetune.py) for details.
 
 
 
diff --git a/train/reproduce/data/README.md b/train/reproduce/data/README.md
new file mode 100644
index 0000000..041966c
--- /dev/null
+++ b/train/reproduce/data/README.md
@@ -0,0 +1,7 @@
+# Data
+
+The Parabank2 dataset can be found here: [https://nlp.jhu.edu/parabank/](https://nlp.jhu.edu/parabank/). Please reformat the data into a `parabank2.json` file like this if you want to use our provided script:
+```
+{"text": "1994 German Grand Prix", "summary": "1994 Grand Prix of Germany"}
+{"text": "2004 OFC Nations Cup", "summary": "2004 Ocean Cup of Nations"}
+```
\ No newline at end of file