col_all.append("label")
// prepare data pack.
df = pd.DataFrame(data, columns=col_all)
df.fillna("missing") // avoid tokenization exception.
// Segment input into 3 dataframes.
relation = df[col_relation]
left = df[["id_left", "text_left"]].drop_duplicates(["id_left"])
left.set_index("id_left", inplace=True)
// Infer the length of the text left
// left["length_left"] = left.apply(lambda r: len(r["text_left"]), axis=1)
After Change
if "id_left" not in df:
id_left = _gen_ids(df, "text_left", "L-")
else:
id_left = df["id_left"]
if "id_right" not in df:
id_right = _gen_ids(df, "text_right", "R-")
else:
id_right = df["id_right"]
