A progress bar implementation for numba functions using tqdm.
The module provides the class ProgressBar
that works as a wrapper around the
tqdm.tqdm
progress bar.
It works by spawning a separate thread that updates the tqdm
progress bar
based on an atomic counter which can be accessed within a numba nopython function.
The progress bar works with parallel as well as sequential numba functions.
pip install numba-progress
git clone https://github.com/mortacious/numba-progress.git
cd numba-progress
python setup.py install
from numba import njit
from numba_progress import ProgressBar
num_iterations = 100
@njit(nogil=True)
def numba_function(num_iterations, progress_proxy):
for i in range(num_iterations):
#<DO CUSTOM WORK HERE>
progress_proxy.update(1)
with ProgressBar(total=num_iterations) as progress:
numba_function(num_iterations, progress)
The ProgressBar
also works within parallel functions out of the box.
from numba import njit, prange
from numba_progress import ProgressBar
num_iterations = 100
@njit(nogil=True, parallel=True)
def numba_function(num_iterations, progress_proxy):
for i in prange(num_iterations):
#<DO CUSTOM WORK HERE>
progress_proxy.update(1)
with ProgressBar(total=num_iterations) as progress:
numba_function(num_iterations, progress)
See also the examples
folder for more usage examples.