Gradient noise defense error

[shanefeng123]

When I tried to use gradient noise as a defense, I encountered the below NotImplementedError.

Running input #0 of 64.
reference: 
========================
 hurts to be you! Sunshine all weekend and you don`t get to play. That`s sad.
========================
748/24/24
L1 Position 0
Traceback (most recent call last):
  File "/home/xinguo/dager-gradient-inversion/attack.py", line 418, in <module>
    main()
  File "/home/xinguo/dager-gradient-inversion/attack.py", line 353, in main
    prediction, reference = reconstruct(args, device, sample, metric, model_wrapper)
                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/xinguo/dager-gradient-inversion/attack.py", line 114, in reconstruct
    res_pos, res_ids, res_types, sentence_ends = filter_l1(args, model_wrapper, R_Qs)
                                                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/xinguo/dager-gradient-inversion/attack.py", line 40, in filter_l1
    raise NotImplementedError
NotImplementedError

This is the command I ran.

python attack.py --dataset sst2 --split val --n_inputs 64 --batch_size 1 --l1_filter maxB --l2_filter non-overlap --model_path bert-base-uncased --task seq_class --cache_dir ./models_cache --defense_noise 0.05