search

facebookresearch / habitat-sim

A flexible, high-performance 3D simulator for Embodied AI research.
https://aihabitat.org/
MIT License
2.58k stars 418 forks source link

Import error in ECCV_2020_Navigation.ipynb #1147

Closed KenaHemnani closed 3 years ago

KenaHemnani commented 3 years ago

In the following colab notebook : https://colab.research.google.com/github/facebookresearch/habitat-sim/blob/master/examples/tutorials/colabs/ECCV_2020_Navigation.ipynb

I am encountering following error in the cell Colab Setup and Imports ,whereas the code was working fine before 3 days. Please, give me suggestions to solve this: RuntimeError Traceback (most recent call last) RuntimeError: module compiled against API version 0xe but this version of numpy is 0xd `ImportError Traceback (most recent call last)

in () 18 from PIL import Image 19 ---> 20 import habitat_sim 21 from habitat_sim.utils import common as utils 22 from habitat_sim.utils import viz_utils as vut 3 frames /usr/local/lib/python3.7/dist-packages/habitat_sim/__init__.py in () 11 if not getattr(builtins, "__HSIM_SETUP__", False): 12 # TODO: kept only for compatibiliy with existing code. Please gradually remove ---> 13 from habitat_sim import ( # noqa: F401 14 agent, 15 attributes, /usr/local/lib/python3.7/dist-packages/habitat_sim/agent/__init__.py in () 7 # type: ignore 8 ----> 9 from .agent import * # noqa: F401,F403 10 from .controls import * # noqa: F401,F403 11 /usr/local/lib/python3.7/dist-packages/habitat_sim/agent/agent.py in () 10 import magnum as mn 11 import numpy as np ---> 12 import quaternion 13 14 import habitat_sim.errors /usr/local/lib/python3.7/dist-packages/quaternion/__init__.py in () 23 import numpy as np 24 ---> 25 from .numpy_quaternion import ( 26 quaternion, _eps, slerp_evaluate, squad_evaluate, 27 # slerp_vectorized, squad_vectorized, slerp, squad, ImportError: numpy.core.multiarray failed to import --------------------------------------------------------------------------- NOTE: If your import is failing due to a missing package, you can manually install dependencies using either !pip or !apt. To view examples of installing some common dependencies, click the "Open Examples" button below. ---------------------------------------------------------------------------`
Skylion007 commented 3 years ago

Yeah, this is an issue with a badly specificied requirements for a new Conda package. I opened an issue: https://github.com/moble/quaternion/issues/161 . If it doesn't get resolved promptly, we will create a workaround.

Until then updating numpy in Colab (with conda) should clear the issue or downgrading quaternion to an earllier release (the later is probably easier.

moble commented 3 years ago

As mentioned on that other issue, this has nothing to do with quaternion; it's just quaternion's bad luck that it is the first thing that tries to actually use numpy after numpy is imported.

This is a fairly common effect of bad installs. In particular, the output shows that the correct (most recent) version of numpy is actually getting installed, but something about python is failing to find the right version of each part of numpy. In this case, I would guess that it's because that notebook tries to install everything, and then use it immediately inside colab. I see, for example, that conda complains about PYTHONPATH being set, which is always a bad idea.

In my testing, it's enough to just restart the runtime (not "factory reset"; just regular restart) after that first installation step. Basically, this is just an inadequate way of installing packages with python.

Skylion007 commented 3 years ago

Yeah, there are some tricks needed install conda packages in Colab without forcing the user to restart. I just pinned the conda version to use the right numpy and it should work for now.

KenaHemnani commented 3 years ago

As mentioned on that other issue, this has nothing to do with quaternion; it's just quaternion's bad luck that it is the first thing that tries to actually use numpy after numpy is imported.

This is a fairly common effect of bad installs. In particular, the output shows that the correct (most recent) version of numpy is actually getting installed, but something about python is failing to find the right version of each part of numpy. In this case, I would guess that it's because that notebook tries to install everything, and then use it immediately inside colab. I see, for example, that conda complains about PYTHONPATH being set, which is always a bad idea. In my testing, it's enough to just restart the runtime after that first installation step. Basically, this is just an inadequate way of installing packages with python.

Yeah , Thank you for the quick response. Restarting run time after first installation worked for me . Thanks.

KenaHemnani commented 3 years ago

Yeah, there are some tricks needed install conda packages in Colab without forcing the user to restart. I just pinned the conda version to use the right numpy and it should work for now.

Thank you for the quick response. I just tried the new notebook from the website again from https://colab.research.google.com/github/facebookresearch/habitat-sim/blob/master/examples/tutorials/colabs/ECCV_2020_Navigation.ipynb

But, it is giving same error. May be I didn't understood your solution. Can you elaborate more about it. Thank you.

aclegg3 commented 3 years ago

Confirmed that restarting the runtime after installation is a work-around for this issue.