you should only change the "map" function to "apply" function, as:
df_test.apply(lambda x: x.replace('blah1',np.nan))