Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

preprocess squad in sft format #6727

Merged
merged 2 commits into from
May 25, 2023
Merged

preprocess squad in sft format #6727

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
df8f21b
preprocess squad in sft format
arendu May 25, 2023
2bc1e91
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] May 25, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
36 changes: 23 additions & 13 deletions scripts/dataset_processing/nlp/squad/prompt_learning_squad_preprocessing.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,6 +56,7 @@ def main():
parser.add_argument("--save-name-base", type=str, default="squad")
parser.add_argument("--include-topic-name", action='store_true')
parser.add_argument("--random-seed", type=int, default=1234)
parser.add_argument("--sft-format", action='store_true')
args = parser.parse_args()

train_data_dict = json.load(open(f"{args.data_dir}/{args.train_file}"))
Expand All @@ -65,21 +66,21 @@ def main():

save_name_base = f"{args.data_dir}/{args.save_name_base}"

process_data(train_data, val_data, save_name_base, args.include_topic_name)
process_data(train_data, val_data, save_name_base, args.include_topic_name, args.sft_format)


def process_data(train_data, val_data, save_name_base, include_topic):
train_set = extract_questions(train_data, include_topic, split="train")
val_set = extract_questions(val_data, include_topic, split="val")
test_set = extract_questions(val_data, include_topic, split="test")
def process_data(train_data, val_data, save_name_base, include_topic, sft_format):
train_set = extract_questions(train_data, include_topic, sft_format, split="train")
val_set = extract_questions(val_data, include_topic, sft_format, split="val")
test_set = extract_questions(val_data, include_topic, sft_format, split="test")

gen_file(train_set, save_name_base, 'train')
gen_file(val_set, save_name_base, 'val')
gen_file(test_set, save_name_base, 'test', make_ground_truth=True)
gen_file(test_set, save_name_base, 'test', make_ground_truth=False)
gen_file(train_set, save_name_base, 'train', sft_format)
gen_file(val_set, save_name_base, 'val', sft_format)
gen_file(test_set, save_name_base, 'test', sft_format, make_ground_truth=True)
gen_file(test_set, save_name_base, 'test', sft_format, make_ground_truth=False)


def extract_questions(data, include_topic, split):
def extract_questions(data, include_topic, sft_format, split):
processed_data = []

# Iterate over topics, want to keep them seprate in train/val/test splits
Expand Down Expand Up @@ -109,7 +110,13 @@ def extract_questions(data, include_topic, split):
except IndexError:
continue

example_json = {"taskname": "squad", "context": context, "question": question, "answer": answers}
if sft_format:
example_json = {
"input": f"User: Context:{context} Question:{question}\n\nAssistant:",
"output": answers,
}
else:
example_json = {"taskname": "squad", "context": context, "question": question, "answer": answers}

if include_topic:
example_json["topic"] = topic
Expand All @@ -120,7 +127,7 @@ def extract_questions(data, include_topic, split):
return processed_data


def gen_file(data, save_name_base, split_type, make_ground_truth=False):
def gen_file(data, save_name_base, split_type, sft_format, make_ground_truth=False):
save_path = f"{save_name_base}_{split_type}.jsonl"

if make_ground_truth:
Expand All @@ -133,7 +140,10 @@ def gen_file(data, save_name_base, split_type, make_ground_truth=False):

# Dont want labels in the test set
if split_type == "test" and not make_ground_truth:
del example_json["answer"]
if sft_format:
example_json["output"] = ""
else:
del example_json["answer"]

save_file.write(json.dumps(example_json) + '\n')

Expand Down