ORIG_DATASET_PATH="../1_raw_dataset/Should-It-Be-Executed-Or-Processed/datasets/train_dataset.json"
OUT_PATH="sep_head_ident.jsonl"
import json
datas = json.loads(open(ORIG_DATASET_PATH).read())
out_file = open(OUT_PATH,"w")
def craft_get_data(data):
inst = data["system_prompt"]
data = data["data_prompt_clean"]
if len(data.split(" ")) < 3:
return
inst_data = f"{inst} {data}"
inv_inst_data = f"{data} {inst}"
out_file.write(json.dumps([inst_data,inv_inst_data]) + "\n")
for data in datas:
craft_get_data(data)