String similarity package in cudF for fuzzy merge

I am trying to do a fuzzy merge of 2 files and have tried to run this code with the new cudf.pandas but getting error regarding the availability of the rapid fuzz package even after installation.

%load_ext cudf.pandas
import pandas as pd
from rapidfuzz import fuzz, process

def fuzzy_merge(df1, df2, key1, key2, threshold=90, limit=1):
“”"
Perform a fuzzy merge and return both matching and unmatching DataFrames.
“”"
s = df2[key2].tolist()
m = df1[key1].apply(lambda x: process.extract(x, s, scorer=fuzz.partial_ratio, limit=limit))
df1[‘matches’] = m

I then tried this:
import cudf
import cupy as cp
from cudf.utils.string_similarity import levenshtein_distance ( and also cudf.levenshtein_distance)

def gpu_fuzzy_merge(df1, df2, key1, key2, threshold=90, limit=1):
“”"
Perform a fuzzy merge using GPU acceleration.
“”"
# Convert pandas DataFrames to cuDF DataFrames if they aren’t already
if not isinstance(df1, cudf.DataFrame):
df1 = cudf.DataFrame.from_pandas(df1)
if not isinstance(df2, cudf.DataFrame):
df2 = cudf.DataFrame.from_pandas(df2) etc.

Still getting errors.

Is there any GPU accelerated package that can be used similar to the rapidfuzz and levenshtein_distance for fuzzy merge?