NotImplementedError: Relevance propagation not implemented for layer type <class 'models.common.BottleneckCSP'>

[HansYeoh]

Hi @akarasman ,
I plan to reproduce your code, but when I run the sample command I get a first off error as below. Since I am not very familiar with the yolo model, I would like to ask you for advice.
Error:

$python3 explain.py --source=data/images/me.png --weights=yolov5s.pt --explain-class='person'
explain: weights=['yolov5s.pt'], source=data/images/me.png, imgsz=[640, 640], conf_thres=0.25, iou_thres=0.45, max_det=1000, device=, view_img=False, save_txt=False, save_conf=False, save_crop=False, nosave=False, classes=None, agnostic_nms=False, visualize=False, update=False, project=runs/explain, name=exp, exist_ok=False, line_thickness=3, hide_labels=False, hide_conf=False, half=False, dnn=False, power=1, contrastive=False, b1=1.0, b2=1.0, explain_class=person, conf=False, max_class_only=True, box_xywh=None, smooth_ks=1, box_xyxy=None, cmap=magma
YOLOv5 🚀 2022-8-1 torch 1.10.1+cu111 CUDA:0 (NVIDIA GeForce RTX 3070, 7971MiB)

Fusing layers... 
Model Summary: 232 layers, 7459581 parameters, 0 gradients
Traceback (most recent call last):
  File "/home/why/Desktop/LRP/inverter.py", line 199, in invert
    return self.inv_funcs[type(layer)](self, layer, relevance, **kwargs) 
KeyError: <class 'models.common.BottleneckCSP'>

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "explain.py", line 391, in <module>
    main(opt)
  File "explain.py", line 387, in main
    run(**vars(opt))
  File "/home/why/anaconda3/envs/lrp/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
    return func(*args, **kwargs)
  File "explain.py", line 282, in run
    get_explanation(inn_model, init, contrastive, b1, b2, int(cls),
  File "explain.py", line 54, in get_explanation
    lrp_out = inn_model.innvestigate(in_tensor=None, initializer=init)
  File "/home/why/Desktop/LRP/innvestigator.py", line 351, in innvestigate
    relevance = self.inverter(layer, relevance)
  File "/home/why/Desktop/LRP/inverter.py", line 206, in __call__
    return self.invert(layer, relevance, **kwargs)
  File "/home/why/Desktop/LRP/inverter.py", line 201, in invert
    raise NotImplementedError(f'Relevance propagation not implemented for layer type {type(layer)}')
NotImplementedError: Relevance propagation not implemented for layer type <class 'models.common.BottleneckCSP'>

And here's the list of dependencies

$ conda list
# packages in environment at /home/why/anaconda3/envs/lrp:
#
# Name                    Version                   Build  Channel
_libgcc_mutex             0.1                        main    defaults
absl-py                   2.0.0                     <pip>
asttokens                 2.4.0                     <pip>
backcall                  0.2.0                     <pip>
Bottleneck                1.3.7                     <pip>
ca-certificates           2023.08.22           h06a4308_0    defaults
cachetools                5.3.1                     <pip>
certifi                   2023.7.22                 <pip>
charset-normalizer        3.3.0                     <pip>
contourpy                 1.1.1                     <pip>
cycler                    0.12.1                    <pip>
decorator                 5.1.1                     <pip>
executing                 2.0.0                     <pip>
fonttools                 4.43.1                    <pip>
google-auth               2.23.3                    <pip>
google-auth-oauthlib      1.0.0                     <pip>
grpcio                    1.59.0                    <pip>
idna                      3.4                       <pip>
importlib-metadata        6.8.0                     <pip>
importlib-resources       6.1.0                     <pip>
ipython                   8.12.3                    <pip>
jedi                      0.19.1                    <pip>
joblib                    1.3.2                     <pip>
kiwisolver                1.4.5                     <pip>
ld_impl_linux-64          2.38                 h1181459_1    defaults
libffi                    3.3                  he6710b0_2    defaults
libgcc-ng                 9.1.0                hdf63c60_0    defaults
libstdcxx-ng              9.1.0                hdf63c60_0    defaults
Markdown                  3.5                       <pip>
MarkupSafe                2.1.3                     <pip>
matplotlib                3.7.3                     <pip>
matplotlib-inline         0.1.6                     <pip>
ncurses                   6.3                  h7f8727e_2    defaults
numpy                     1.24.4                    <pip>
oauthlib                  3.2.2                     <pip>
opencv-python             4.8.1.78                  <pip>
openssl                   1.1.1w               h7f8727e_0    defaults
packaging                 23.2                      <pip>
pandas                    2.0.3                     <pip>
parso                     0.8.3                     <pip>
pexpect                   4.8.0                     <pip>
pickleshare               0.7.5                     <pip>
Pillow                    10.1.0                    <pip>
pip                       23.2.1           py38h06a4308_0    defaults
ply                       3.11                      <pip>
prompt-toolkit            3.0.39                    <pip>
protobuf                  4.24.4                    <pip>
psutil                    5.9.6                     <pip>
ptyprocess                0.7.0                     <pip>
pure-eval                 0.2.2                     <pip>
pyasn1                    0.5.0                     <pip>
pyasn1-modules            0.3.0                     <pip>
Pygments                  2.16.1                    <pip>
pyparsing                 3.1.1                     <pip>
python                    3.8.13               h12debd9_0    defaults
python-dateutil           2.8.2                     <pip>
pytz                      2023.3.post1              <pip>
PyYAML                    6.0.1                     <pip>
readline                  8.1.2                h7f8727e_1    defaults
requests                  2.31.0                    <pip>
requests-oauthlib         1.3.1                     <pip>
rsa                       4.9                       <pip>
scikit-learn              1.3.1                     <pip>
scipy                     1.10.1                    <pip>
seaborn                   0.13.0                    <pip>
setuptools                53.0.0                    <pip>
setuptools                68.0.0           py38h06a4308_0    defaults
sip                       6.7.12                    <pip>
six                       1.16.0                    <pip>
sqlite                    3.38.2               hc218d9a_0    defaults
stack-data                0.6.3                     <pip>
tensorboard               2.14.0                    <pip>
tensorboard-data-server   0.7.1                     <pip>
thop-0.1.1                2209072238                <pip>
threadpoolctl             3.2.0                     <pip>
tk                        8.6.11               h1ccaba5_0    defaults
tomli                     2.0.1                     <pip>
torch                     1.10.1+cu111              <pip>
torchaudio                0.10.1+rocm4.1            <pip>
torchvision               0.11.2+cu111              <pip>
tqdm                      4.66.1                    <pip>
traitlets                 5.11.2                    <pip>
typing_extensions         4.8.0                     <pip>
tzdata                    2023.3                    <pip>
urllib3                   2.0.7                     <pip>
wcwidth                   0.2.8                     <pip>
Werkzeug                  3.0.0                     <pip>
wheel                     0.41.2           py38h06a4308_0    defaults
xz                        5.2.5                h7f8727e_1    defaults
zipp                      3.17.0                    <pip>
zlib                      1.2.11                        0    https://mirrors.sjtug.sjtu.edu.cn/anaconda/pkgs/free