Error in interpreter when the input sequence is more than 1d

[zegnog]

Checklist

• I have verified that the issue exists against the main branch of AllenNLP.
• I have read the relevant section in the contribution guide on reporting bugs.
• I have checked the issues list for similar or identical bug reports.
• I have checked the pull requests list for existing proposed fixes.
• I have checked the CHANGELOG and the commit log to find out if the bug was already fixed in the main branch.
• I have included in the "Description" section below a traceback from any exceptions related to this bug.
• I have included in the "Related issues or possible duplicates" section below all related issues and possible duplicate issues (If there are none, check this box anyway).
• I have included in the "Environment" section below the name of the operating system and Python version that I was using when I discovered this bug.
• I have included in the "Environment" section below the output of pip freeze.
• I have included in the "Steps to reproduce" section below a minimally reproducible example.

Description

In integrated_gradient.py, the gradient is normed assuming the input is of shape (b, seq_len, emb_size...), but sometimes the input is more than one sequence. It will trigger the following error:

Python traceback:

<ipython-input-2-ba1db62dbe43> in <listcomp>(.0)
      2     embedding_grad = numpy.sum(grad[0], axis=1)
      3     norm = numpy.linalg.norm(embedding_grad, ord=1)
----> 4     normalized_grad = [math.fabs(e) / norm for e in embedding_grad]
      5     return normalized_grad

TypeError: only size-1 arrays can be converted to Python scalars

Related issues or possible duplicates

• None

Environment

OS: MacOS

Python version: 3.8.3

Output of pip freeze:

absl-py==0.11.0
allennlp==2.5.0
allennlp-models==2.5.0
altair==4.1.0
appnope @ file:///opt/concourse/worker/volumes/live/0291c9e1-4b15-459f-623e-2770f55be269/volume/appnope_1594338395037/work
argon2-cffi==20.1.0
astor==0.8.1
astroid @ file:///opt/concourse/worker/volumes/live/21fd14a9-2a7e-484b-7394-5a9912cdcf80/volume/astroid_1592498459180/work
astunparse==1.6.3
async-generator==1.10
attrs==20.2.0
autopep8 @ file:///tmp/build/80754af9/autopep8_1596578164842/work
backcall==0.2.0
backports.csv==1.0.7
base58==2.0.1
beautifulsoup4==4.9.3
bidict==0.21.2
bleach==3.2.1
blinker==1.4
blis==0.4.1
bokeh @ file:///opt/concourse/worker/volumes/live/00176cca-b1c3-4fe7-4da0-2ea50fa27756/volume/bokeh_1599056223916/work
boto3==1.15.16
botocore==1.18.16
brotlipy==0.7.0
cachetools==4.2.0
catalogue==1.0.0
certifi==2020.12.5
cffi @ file:///opt/concourse/worker/volumes/live/b9607b09-b777-4ff7-53dc-287727eb8574/volume/cffi_1600699191154/work
chardet==4.0.0
checklist==0.0.11
cheroot==8.5.2
CherryPy==18.6.0
click==7.1.2
cloudpickle @ file:///tmp/build/80754af9/cloudpickle_1598884132938/work
conda==4.9.2
conda-package-handling==1.7.0+0.g7c4a471.dirty
configparser==5.0.2
conllu==4.4
cryptography @ file:///opt/concourse/worker/volumes/live/aeb63a26-659e-4edb-5405-74ba8e0c76f2/volume/cryptography_1601046839724/work
cycler==0.10.0
cymem==2.0.3
cytoolz==0.11.0
dask @ file:///tmp/build/80754af9/dask-core_1600699676946/work
dataclasses==0.6
decorator==4.4.2
defusedxml==0.6.0
dill==0.3.3
distributed @ file:///opt/concourse/worker/volumes/live/88b24afb-2456-43bf-56b4-38902ff2143a/volume/distributed_1601910790507/work
docker-pycreds==0.4.0
email-reply-parser==0.5.12
en-core-web-md @ https://github.com/explosion/spacy-models/releases/download/en_core_web_md-2.3.1/en_core_web_md-2.3.1.tar.gz
en-core-web-sm @ https://github.com/explosion/spacy-models/releases/download/en_core_web_sm-2.3.1/en_core_web_sm-2.3.1.tar.gz
entrypoints==0.3
et-xmlfile==1.0.1
feedparser==6.0.8
filelock==3.0.12
flake8 @ file:///tmp/build/80754af9/flake8_1601911421857/work
Flask==1.1.2
Flask-SocketIO==5.0.1
flatbuffers==1.12
fsspec @ file:///tmp/build/80754af9/fsspec_1597944003862/work
ftfy==5.8
future==0.18.2
gast==0.3.3
gensim==3.8.3
gitdb==4.0.5
GitPython==3.1.11
google-api-core==1.30.0
google-auth==1.32.1
google-auth-oauthlib==0.4.2
google-cloud-core==1.7.1
google-cloud-storage==1.38.0
google-crc32c==1.1.2
google-pasta==0.2.0
google-resumable-media==1.3.1
googleapis-common-protos==1.53.0
greenlet==1.0.0
grpcio==1.34.0
h5py==2.10.0
HeapDict==1.0.1
huggingface-hub==0.0.13
idna==2.10
importlib-metadata @ file:///tmp/build/80754af9/importlib-metadata_1602276842396/work
iniconfig==1.0.1
ipaddress==1.0.23
ipykernel @ file:///opt/concourse/worker/volumes/live/88f541d3-5a27-498f-7391-f2e50ca36560/volume/ipykernel_1596206680118/work/dist/ipykernel-5.3.4-py3-none-any.whl
ipython @ file:///opt/concourse/worker/volumes/live/90527372-a5f7-4d48-68c2-a99de2ba31c7/volume/ipython_1599056199422/work
ipython-genutils==0.2.0
ipywidgets==7.5.1
iso-639==0.4.5
isort @ file:///opt/concourse/worker/volumes/live/97586196-dd0e-4a3f-4d06-330d40d66119/volume/isort_1601490198153/work
itsdangerous==1.1.0
jaraco.classes==3.2.1
jaraco.collections==3.3.0
jaraco.functools==3.3.0
jaraco.text==3.5.0
jdcal==1.4.1
jedi @ file:///opt/concourse/worker/volumes/live/c63c70ef-654a-4d26-6b78-37aef096f225/volume/jedi_1596490811751/work
Jinja2==2.11.2
jmespath==0.10.0
joblib==0.16.0
jsonnet==0.16.0
jsonpickle==1.4.1
jsonschema==3.2.0
jupyter==1.0.0
jupyter-client @ file:///tmp/build/80754af9/jupyter_client_1601311786391/work
jupyter-console==6.4.0
jupyter-core==4.6.3
jupyterlab-pygments==0.1.2
Keras-Preprocessing==1.1.2
kiwisolver==1.2.0
langdetect==1.0.7
lazy-object-proxy==1.4.3
lmdb==1.1.1
locket==0.2.0
lxml==4.6.3
mail-parser==3.12.0
Markdown==3.3.3
MarkupSafe @ file:///opt/concourse/worker/volumes/live/cb778296-98db-45ad-411e-6f726e102dc3/volume/markupsafe_1594371638608/work
matplotlib==3.3.2
mccabe==0.6.1
mistune==0.8.4
mittens==0.2
mkl-fft==1.2.0
mkl-random==1.1.1
mkl-service==2.3.0
more-itertools==8.7.0
msgpack==1.0.0
munch==2.5.0
murmurhash==1.0.2
nbclient==0.5.1
nbconvert==6.0.7
nbformat==5.0.8
neovim==0.3.1
nest-asyncio==1.4.3
nltk==3.5
notebook==6.1.5
numpy==1.19.4
oauthlib==3.1.0
olefile==0.46
openpyxl @ file:///tmp/build/80754af9/openpyxl_1610651698508/work
opt-einsum==3.3.0
overrides==3.1.0
packaging==20.9
pandas==1.2.2
pandocfilters==1.4.3
parso==0.7.0
partd==1.1.0
pathtools==0.1.2
patternfork-nosql==3.6
pdfminer.six==20201018
pexpect @ file:///opt/concourse/worker/volumes/live/8701bb20-ad87-46c7-5108-30c178cf97e5/volume/pexpect_1594383388344/work
pickleshare @ file:///opt/concourse/worker/volumes/live/93ec39d8-05bb-4f84-7efc-98735bc39b70/volume/pickleshare_1594384101884/work
Pillow @ file:///opt/concourse/worker/volumes/live/be1e8a56-c4be-4ffe-4fa7-5a0e9c460b1a/volume/pillow_1594307312933/work
plac==1.1.3
pluggy==0.13.1
portend==2.7.1
preshed==3.0.2
prettyprint==0.1.5
prometheus-client==0.9.0
promise==2.3
prompt-toolkit @ file:///tmp/build/80754af9/prompt-toolkit_1598885458782/work
protobuf==3.14.0
psutil @ file:///opt/concourse/worker/volumes/live/ff72f822-991c-4030-4f3a-8c41d3ac4e4f/volume/psutil_1598370232375/work
ptyprocess==0.6.0
py==1.9.0
py-rouge==1.1
pyarrow==2.0.0
pyasn1==0.4.8
pyasn1-modules==0.2.8
pycodestyle==2.6.0
pycosat==0.6.3
pycparser @ file:///tmp/build/80754af9/pycparser_1594388511720/work
pydeck==0.5.0
pyeasyga==0.3.1
pyflakes==2.2.0
Pygments @ file:///tmp/build/80754af9/pygments_1600458456400/work
pylint @ file:///opt/concourse/worker/volumes/live/ed0164b6-bcc7-4f6b-7dd4-ad89660b5dcb/volume/pylint_1598624018129/work
pynvim==0.4.2
pyOpenSSL @ file:///tmp/build/80754af9/pyopenssl_1594392929924/work
pyparsing==2.4.7
pyrsistent==0.17.3
PySocks @ file:///opt/concourse/worker/volumes/live/85a5b906-0e08-41d9-6f59-084cee4e9492/volume/pysocks_1594394636991/work
pyswarms==1.2.0
pytest==6.1.1
python-crfsuite==0.9.7
python-dateutil==2.8.1
python-docx==0.8.11
python-engineio==4.0.0
python-socketio==5.0.4
pytz==2020.1
PyYAML==5.3.1
pyzmq==19.0.2
qtconsole==5.1.1
QtPy==1.9.0
QuantLib==1.20
regex==2020.9.27
requests==2.25.1
requests-oauthlib==1.3.0
rsa==4.6
ruamel-yaml==0.15.87
s3transfer==0.3.3
sacremoses==0.0.43
scikit-learn==0.23.2
scipy==1.5.2
seaborn==0.11.1
Send2Trash==1.5.0
sentencepiece==0.1.91
sentry-sdk==1.1.0
sgmllib3k==1.0.0
shortuuid==1.0.1
simplejson==3.17.0
six @ file:///opt/concourse/worker/volumes/live/5b31cb27-1e37-4ca5-6e9f-86246eb206d2/volume/six_1605205320872/work
sklearn-crfsuite==0.3.6
smart-open==4.0.1
smmap==3.0.4
sortedcontainers==2.2.2
soupsieve==2.2.1
spacy==2.3.2
srsly==1.0.2
streamlit==0.83.0
subprocess32==3.5.4
tabulate==0.8.7
tblib @ file:///tmp/build/80754af9/tblib_1597928476713/work
tempora==4.1.1
tensorboard==2.4.0
tensorboard-plugin-wit==1.7.0
tensorboardX==2.1
tensorflow-estimator==2.4.0
termcolor==1.1.0
terminado==0.9.1
testpath==0.4.4
textblob==0.15.3
thinc==7.4.1
threadpoolctl==2.1.0
tokenizers==0.10.1
toml @ file:///tmp/build/80754af9/toml_1592853716807/work
toolz @ file:///tmp/build/80754af9/toolz_1601054250827/work
torch==1.7.1
torchtext==0.7.0
torchvision==0.8.2
tornado==6.0.4
tqdm @ file:///tmp/build/80754af9/tqdm_1601923723745/work
traitlets @ file:///tmp/build/80754af9/traitlets_1600712679583/work
transformers==4.3.3
typing-extensions @ file:///tmp/build/80754af9/typing_extensions_1598376058250/work
tzlocal==2.1
urllib3==1.25.11
validators==0.18.2
wandb==0.10.33
wasabi==0.8.0
watchdog==1.0.2
wcwidth @ file:///tmp/build/80754af9/wcwidth_1593447189090/work
webencodings==0.5.1
Werkzeug==1.0.1
widgetsnbextension==3.5.1
word2number==1.1
wrapt==1.12.1
xlrd @ file:///tmp/build/80754af9/xlrd_1608072521494/work
zc.lockfile==2.0
zict==2.0.0
zipp @ file:///tmp/build/80754af9/zipp_1604001098328/work

Steps to reproduce

run the following code, the old one is current way and the new one is proposed new way.

Example source:

# To add a new cell, type '# %%'
# To add a new markdown cell, type '# %% [markdown]'
# %%
import numpy
import math


# %%
def old_norm_grad(grad:numpy.ndarray):
    embedding_grad = numpy.sum(grad[0], axis=1)
    norm = numpy.linalg.norm(embedding_grad, ord=1)
    normalized_grad = [math.fabs(e) / norm for e in embedding_grad]
    return normalized_grad


# %%
def new_norm_grad(grad: numpy.ndarray):
    embedding_grad = numpy.sum(grad[0], axis=-1)
    norm = numpy.linalg.norm(embedding_grad, ord=1, keepdims=True)
    normalized_grad = embedding_grad / norm
    return normalized_grad


# %%
test_grad = numpy.random.rand(1,13,100)
numpy.testing.assert_array_equal(old_norm_grad(test_grad), new_norm_grad(test_grad)) # No error raised


# %%
test_grad = numpy.random.rand(1,2,13,100)
numpy.testing.assert_array_equal(old_norm_grad(test_grad), new_norm_grad(test_grad)) # raised TypeError: only size-1 arrays can be converted to Python scalars


# %%

I have created a commit targeting this issue, if applicable I could pr anytime:
zegnog@2415cb9

[epwalsh]

Hi @zegnog, yes please submit a PR when you get a chance.

[github-actions]

This issue is being closed due to lack of activity. If you think it still needs to be addressed, please comment on this thread 👇