running librosa & numba on raspberry pi 3

问题:

I am trying to run librosa on my raspberry pi 3. After hours of searching through the internet I was finally able to install it but it still throws an error when I try to import it. First, I had problems to install the dependency llvmlite. I finally installed it with the following code: conda install -c numba llvmlite  I use python 3.4 build with miniconda.

After llvmlite was installed I was able to install librosa with pip (not possible with conda)

pi@raspberrypi:~ $ pip install librosa
Collecting librosa
Using cached https://www.piwheels.hostedpi.com/simple/librosa/librosa-
0.5.1-py3-none-any.whl
Requirement already satisfied: resampy>=0.1.2 in 
./miniconda3/lib/python3.4/site-packages (from librosa)
Requirement already satisfied: scipy>=0.13.0 in 
...
Installing collected packages: librosa
Successfully installed librosa-0.5.1

Apparently librosa is installed.

pi@raspberrypi:~ $ pip list
DEPRECATION: The default format will switch to columns in the future. 
You can use --format=(legacy|columns) (or define a format=
(legacy|columns) in your pip.conf under the [list] section) to disable 
this warning.
anaconda-client (1.0.2)
audioread (2.1.5)
awscli (1.11.170)
boto3 (1.4.7)
botocore (1.7.28)
clyent (0.4.0)
colorama (0.3.7)
conda (3.16.0)
conda-env (2.4.2)
Cython (0.27.1)
decorator (4.1.2)
docutils (0.14)
jmespath (0.9.3)
joblib (0.11)
librosa (0.5.1)
llvmlite (0.7.0.dev0+21.gcda19bd.dirty)
mock (2.0.0)
numba (0.35.0)
numpy (1.13.3)
pbr (3.1.1)
Pillow (2.9.0)
pip (9.0.1)
protobuf (3.4.0)
pyasn1 (0.3.7)
PyAudio (0.2.11)
pycosat (0.6.1)
pycrypto (2.6.1)    
python-dateutil (2.6.1)
pytz (2015.4)
PyYAML (3.12)
requests (2.7.0)
resampy (0.2.0)
rsa (3.4.2)
s3transfer (0.1.11)
scikit-learn (0.18.2)
scipy (0.19.1)
setuptools (18.1)
six (1.11.0)
tensorflow (1.1.0)
Werkzeug (0.12.2)
wheel (0.30.0)

Trying to import librosa throws the following error though:

pi@raspberrypi:~ $ python
Python 3.4.3 |Continuum Analytics, Inc.| (default, Aug 21 2015, 00:53:08) 
[GCC 4.6.3] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import librosa
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/pi/miniconda3/lib/python3.4/site-packages/librosa/__init__.py", line 12, in <module>
    from . import core
  File "/home/pi/miniconda3/lib/python3.4/site-packages/librosa/core/__init__.py", line 108, in <module>
    from .time_frequency import *  # pylint: disable=wildcard-import
  File "/home/pi/miniconda3/lib/python3.4/site-packages/librosa/core/time_frequency.py", line 10, in <module>
    from ..util.exceptions import ParameterError
  File "/home/pi/miniconda3/lib/python3.4/site-packages/librosa/util/__init__.py", line 70, in <module>
    from . import decorators
  File "/home/pi/miniconda3/lib/python3.4/site-packages/librosa/util/decorators.py", line 67, in <module>
    from numba.decorators import jit as optional_jit
  File "/home/pi/miniconda3/lib/python3.4/site-packages/numba/__init__.py", line 12, in <module>
    from .special import typeof, prange
  File "/home/pi/miniconda3/lib/python3.4/site-packages/numba/special.py", line 4, in <module>
    from .parfor import prange
  File "/home/pi/miniconda3/lib/python3.4/site-packages/numba/parfor.py", line 20, in <module>
    from numba import array_analysis, postproc, typeinfer
  File "/home/pi/miniconda3/lib/python3.4/site-packages/numba/array_analysis.py", line 9, in <module>
    from numba import ir, analysis, types, config, cgutils, typing
  File "/home/pi/miniconda3/lib/python3.4/site-packages/numba/cgutils.py", line 22, in <module>
    true_bit = bool_t(1)
TypeError: 'IntType' object is not callable
>>> 

Trying to import numba throws the following error:

pi@raspberrypi:~ $ python
Python 3.4.3 |Continuum Analytics, Inc.| (default, Aug 21 2015, 00:53:08) 
[GCC 4.6.3] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import numba
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/pi/miniconda3/lib/python3.4/site-packages/numba/__init__.py", line 12, in <module>
    from .special import typeof, prange
  File "/home/pi/miniconda3/lib/python3.4/site-packages/numba/special.py", line 4, in <module>
    from .parfor import prange
  File "/home/pi/miniconda3/lib/python3.4/site-packages/numba/parfor.py", line 20, in <module>
    from numba import array_analysis, postproc, typeinfer
  File "/home/pi/miniconda3/lib/python3.4/site-packages/numba/array_analysis.py", line 9, in <module>
    from numba import ir, analysis, types, config, cgutils, typing
  File "/home/pi/miniconda3/lib/python3.4/site-packages/numba/cgutils.py", line 22, in <module>
    true_bit = bool_t(1)
TypeError: 'IntType' object is not callable
>>> 

Has anyone a clue what the problem could be? I am grateful for any hints or tips. Thank you!!!!!

回答1:

As of writing, the latest Raspbian release has llvm-3.8, so per this github comment you need to install llvmlite v0.15.0 with numba v0.30.1. If you're on a different llvm then you will need to explore what versions of llvmlite and numba to use with that version of llvm. Here are the commands that got me set up to get a successful import librosa:

sudo apt install libblas-dev llvm python3-pip python3-scipy
virtualenv --system-site-packages -p python3 env
source env/bin/activate
pip install llvmlite==0.15.0
pip install numba==0.30.1
pip install librosa

Note I'm using virtualenv and dragging in scipy and numpy from the system packages, otherwise it's hard to get the fortran & c dependencies right plus those take ages to compile on a RPi. If you want to install system-wide then drop the 2nd and 3rd line and put sudo in front of the pip commands. I've also frozen my requirements in this requirements.txt file so if you download that then you can run a single command pip install -r requirements.txt

回答2:

Thanks @MatthewBerryman, you got me over the hump! On the newest Raspian release (stretch) I was successful with the following after several hours of frustration of trying to get librosa installed on Raspian jessie (which my RPi3 came with). Having said this, the following procedure may also work with jessie.

First, update your system's package list and upgrade all your installed packages to their latest versions with the command:

sudo apt-get update
sudo apt-get dist-upgrade

Install Python science stack:

sudo pip3 install numpy --upgrade 
sudo apt-get install python3-pandas

(Also seems to install matplotlib, scipy)

sudo apt-get install python3-sklearn

Then, install the low-level virtual machine, LLVM (per @MatthewBerryman, I used llvm 3.8 and llvmlite 0.15.0, and not the newest combination where I couldn't find the packages.) After installing llvm-3.8, a symbolic link needs to be defined before installing llvmlite.

sudo apt-get install llvm-3.8
sudo ln -s /usr/bin/llvm-config-3.8 /usr/bin/llvm-config
sudo pip3 install llvmlite==0.15.0
sudo pip3 install numba==0.32.0

Numba is 0.32.0 because if it's the newest (0.36), it will not import because of an llvm mismatch, and if it's a lower version, the librosa install will upgrade it to the newest version.

Finally, install librosa:

sudo pip3 install librosa

However, when trying to import librosa, it still throws and error, namely

ImportError: libf77blas.so.3: cannot open shared object file: No such file or directory

Googling this error indicated this would fix it:

sudo apt-get install libatlas-base-dev

And it did; however, I have no idea why.

To summarize, this procedure installs librosa, and there is no error when trying this:

...$ python3
Python 3.5.3 (default, Jan 19 2017, 14:11:04)
...
>>>import librosa
>>>

回答3:

sudo pip install librosa==0.4.2 worked for me. There was warning yet works fine on Raspberry pi3 (OS:raspbian-jessie)