Kernel dies importing spacy

[bosanipietro]

How to reproduce the behaviour

import spacy

Your Environment

• Operating System: macOS with Apple M1 chip
• Python Version Used: 3.9.4
• spaCy Version Used: 3.0.0
• Environment Information:

Importing spaCy library into Jupiter notebook within the above environment kills my kernel, with the package failing to import. Any help would be greatly appreciated, thanks for all your hard work!

[polm]

Sorry you're having trouble with that. We have had reports of issues with M1 chips, and in general Python and pip had trouble adjusting immediately after they were released, though at this point I think most people are able to use it without issue.

1. Can you upgrade spaCy to the most recent version and try again?
2. Do you not get any error output? If you do can you share it? How do you know your kernel died?
3. Can you import spaCy in Python on the command line?

[abhi-vellala]

@polm You mean which app should I choose to use Rosetta? Anaconda? I've tried to reinstall spacy. It still crashes the kernel.

[polm]

@abhi-vellala If you are getting Illegal instruction: 4, the issue is that your Python and the package you're trying to use have different CPU architectures. You might be able to fix that in Rosetta settings as indicated above. If that doesn't work, please answer the questions above so we have information to help you - just knowing the kernels crashes doesn't give us enough information to figure out what's wrong.

[isobarbaric]

similar issue, I couldn't get python -m spacy download en_core_web_sm to work normally, so I tried with Rosetta, but to no avail; the error I keep getting is this: [1] (a random number here) illegal hardware instruction python -m spacy download en_core_web_sm (the number after the [1] is different every time)

[rohitsahayy]

i am having the same problem . The kernel dies when trying to import Spacy.
Operating System: macOS with Apple M2 chip
Python Version Used: 3.9.12
spaCy Version Used: 3.5.3

[svlandeg]

Please open a new issue if you run into this, and share all information as outlined in the section "How to ask a good question" here. Thanks!

[wentingd17]

I had the same issue before.
I tried to create a conda Env, install spacy packages there and add it to my jupyter notebook. Then run the jupyter notebook in that new conda Env then it works.

[abhi413sh]

@wentingd17 how you did it?

[bobiq11]

I also have this problem. How do I solve this?

[svlandeg]

Can you answer the questions Paul summed up here?

[anammari]

I am encountering the same issue on a Jupyter server deployed as a HuggingFace Space.
It has been working with no issues previously but just today started facing the kernel dies issue

Tried the latest SpaCy version v3.5.3 as well as v3.5.0

Python version is v3.10

Running "!pip show spacy" in Jupyter works fine and produces the following output:
Name: spacy
Version: 3.5.0
Summary: Industrial-strength Natural Language Processing (NLP) in Python
Home-page: https://spacy.io/
Author: Explosion
Author-email: [email protected]
License: MIT
Location: /home/jovyan/.local/lib/python3.10/site-packages
Requires: catalogue, cymem, jinja2, langcodes, murmurhash, numpy, packaging, pathy, preshed, pydantic, requests, setuptools, smart-open, spacy-legacy, spacy-loggers, srsly, thinc, tqdm, typer, wasabi
Required-by:

[vinbo8]

Could you open a new thread for this issue, with server logs instead of images if you have access to those?

[anammari]

Thanks @vin-ivar
New issue has been logged #12665
I've added both the HF Space build logs and the Space container run-time logs

[Aisling-Kinsella]

I'm having this same problem when I import spacy into an environment pre-loaded with top2vec. Previous spacy env working fine, top2vec env working fine before spacy install in that environment. Can't find a workaround except to switch environments for relevant work. Is there some way to find out what is causing this problem though, as I've followed the flow in answer above and spacy is updated, installed also via command line into the env but there's no problem with the install, just the dead kernel when the code hits the import spacy command in jupyter notebook

[adrianeboyd]

This is probably due to an interaction with tensorflow. If you uninstall tensorflow, does spacy work again?

spacy (well, thinc) imports tensorflow automatically if it's installed, but users rarely use tensorflow with spacy. For now, the easiest workaround is to use a separate venv for spacy where tensorflow isn't installed.

But this is becoming more of a problem with newer versions of tensorflow, so I think we should modify thinc so that it no longer imports tensorflow by default, only if the user requests it: explosion/thinc#890

[patkovacs]

if i install it with conda, it works:
conda install -c conda-forge spacy