tqdm
means "progress" in Arabic (taqadum, تقدّم) and an abbreviation for "I love you so much" in Spanish (te quiero demasiado).
Instantly make your loops show a smart progress meter - just wrap any iterable with tqdm(iterable)
, and you're done!
76%|████████████████████████████ | 7568/10000 [00:33<00:10, 229.00it/s]
trange(N)
can be also used as a convenient shortcut for tqdm(xrange(N))
.
REPL: ptpython
It can also be executed as a module with pipes:
Overhead is low -- about 60ns per iteration (80ns with tqdm_gui
), and is unit tested against performance regression. By comparison, the well-established ProgressBar has an 800ns/iter overhead.
In addition to its low overhead, tqdm
uses smart algorithms to predict the remaining time and to skip unnecessary iteration displays, which allows for a negligible overhead in most cases.
tqdm
works on any platform (Linux, Windows, Mac, FreeBSD, NetBSD, Solaris/SunOS), in any console or in a GUI, and is also friendly with IPython/Jupyter notebooks.
tqdm
does not require any dependencies (not even curses
!), just Python and an environment supporting carriage return \r
and line feed \n
control characters.
Table of contents
Pull and install in the current directory:
The list of all changes is available either on GitHub's Releases: , on the wiki or on crawlers such as allmychanges.com.
tqdm
is very versatile and can be used in a number of ways. The three main ones are given below.
Wrap tqdm()
around any iterable:
trange(i)
is a special optimised instance of tqdm(range(i))
:
Instantiation outside of the loop allows for manual control over tqdm()
:
Manual control on tqdm()
updates by using a with
statement:
If the optional variable total
(or an iterable with len()
) is provided, predictive stats are displayed.
with
is also optional (you can just assign tqdm()
to a variable, but in this case don't forget to del
or close()
at the end:
Perhaps the most wonderful use of tqdm
is in a script or on the command line. Simply inserting tqdm
(or python -m tqdm
) between pipes will pass through all stdin
to stdout
while printing progress to stderr
.
The example below demonstrated counting the number of lines in all Python files in the current directory, with timing information included.
Note that the usual arguments for tqdm
can also be specified.
Backing up a large directory?
The most common issues relate to excessive output on multiple lines, instead of a neat one-line progress bar.
- Consoles in general: require support for carriage return (
CR
,\r
). - Nested progress bars:
- Unicode:
- Environments which report that they support unicode will have solid smooth progressbars. The fallback is an ascii-only bar.
- Windows consoles often only partially support unicode and thus often require explicit ascii=True (also here). This is due to either normal-width unicode characters being incorrectly displayed as "wide", or some unicode characters not rendering.
- Wrapping enumerated iterables: use
enumerate(tqdm(...))
instead oftqdm(enumerate(...))
. The same applies tonumpy.ndenumerate
. This is because enumerate functions tend to hide the length of iterables.tqdm
does not. - Wrapping zipped iterables has similar issues due to internal optimisations.
tqdm(zip(a, b))
should be replaced withzip(tqdm(a), b)
or evenzip(tqdm(a), tqdm(b))
.
If you come across any other difficulties, browse and file .
class tqdm(object):
"""
Decorate an iterable object, returning an iterator which acts exactly
like the original iterable, but prints a dynamically updating
progressbar every time a value is requested.
"""
def __init__(self, iterable=None, desc=None, total=None, leave=True,
file=None, ncols=None, mininterval=0.1,
maxinterval=10.0, miniters=None, ascii=None, disable=False,
unit='it', unit_scale=False, dynamic_ncols=False,
smoothing=0.3, bar_format=None, initial=0, position=None,
postfix=None):
- iterable : iterable, optional
Iterable to decorate with a progressbar. Leave blank to manually manage the updates.
- desc : str, optional
Prefix for the progressbar.
- total : int, optional
The number of expected iterations. If (default: None), len(iterable) is used if possible. As a last resort, only basic progress statistics are displayed (no ETA, no progressbar). If
gui
is True and this parameter needs subsequent updating, specify an initial arbitrary large positive integer, e.g. int(9e9).
- leave : bool, optional
If [default: True], keeps all traces of the progressbar upon termination of iteration.
- file :
io.TextIOWrapper
orio.StringIO
, optional Specifies where to output the progress messages (default: sys.stderr). Uses
file.write(str)
andfile.flush()
methods.
- file :
- ncols : int, optional
The width of the entire output message. If specified, dynamically resizes the progressbar to stay within this bound. If unspecified, attempts to use environment width. The fallback is a meter width of 10 and no limit for the counter and statistics. If 0, will not print any meter (only stats).
- mininterval : float, optional
Minimum progress display update interval, in seconds [default: 0.1].
- maxinterval : float, optional
Maximum progress display update interval, in seconds [default: 10]. Automatically adjusts
miniters
to correspond tomininterval
after long display update lag. Only works ifdynamic_miniters
or monitor thread is enabled.
- miniters : int, optional
Minimum progress display update interval, in iterations. If 0 and
dynamic_miniters
, will automatically adjust to equalmininterval
(more CPU efficient, good for tight loops). If > 0, will skip display of specified number of iterations. Tweak this andmininterval
to get very efficient loops. If your progress is erratic with both fast and slow iterations (network, skipping items, etc) you should set miniters=1.
- ascii : bool, optional
If unspecified or False, use unicode (smooth blocks) to fill the meter. The fallback is to use ASCII characters
1-9 #
.
- disable : bool, optional
Whether to disable the entire progressbar wrapper [default: False].
- unit : str, optional
String that will be used to define the unit of each iteration [default: it].
- unit_scale : bool or int or float, optional
If 1 or True, the number of iterations will be reduced/scaled automatically and a metric prefix following the International System of Units standard will be added (kilo, mega, etc.) [default: False]. If any other non-zero number, will scale total and n.
- dynamic_ncols : bool, optional
If set, constantly alters
ncols
to the environment (allowing for window resizes) [default: False].
- smoothing : float, optional
Exponential moving average smoothing factor for speed estimates (ignored in GUI mode). Ranges from 0 (average speed) to 1 (current/instantaneous speed) [default: 0.3].
- bar_format : str, optional
Specify a custom bar string formatting. May impact performance. [default: '{l_bar}{bar}{r_bar}'], where l_bar='{desc}: {percentage:3.0f}% {n_fmt}/{total_fmt} [{elapsed}<{remaining}, ' '{rate_fmt}{postfix}]' Possible vars: l_bar, bar, r_bar, n, n_fmt, total, total_fmt, percentage, rate, rate_fmt, rate_noinv, rate_noinv_fmt, rate_inv, rate_inv_fmt, elapsed, remaining, desc, postfix. Note that a trailing ": " is automatically removed after {desc} if the latter is empty.
- initial : int, optional
The initial counter value. Useful when restarting a progress bar [default: 0].
- position : int, optional
Specify the line offset to print this bar (starting from 0) Automatic if unspecified. Useful to manage multiple bars at once (eg, from threads).
- postfix : dict or
*
, optional Specify additional stats to display at the end of the bar. Calls
set_postfix(**postfix)
if possible (dict).
- postfix : dict or
- unit_divisor : float, optional
[default: 1000], ignored unless unit_scale is True.
- delim : chr, optional
Delimiting character [default: 'n']. Use '0' for null. N.B.: on Windows systems, Python converts 'n' to 'rn'.
- buf_size : int, optional
String buffer size in bytes [default: 256] used when
delim
is specified.
- bytes : bool, optional
If true, will count bytes and ignore
delim
.
- out : decorated iterator.
def update(self, n=1):
"""
Manually update the progress bar, useful for streams
such as reading files.
E.g.:
>>> t = tqdm(total=filesize) # Initialise
>>> for current_buffer in stream:
... ...
... t.update(len(current_buffer))
>>> t.close()
The last line is highly recommended, but possibly not necessary if
``t.update()`` will be called in such a way that ``filesize`` will be
exactly reached and printed.
Parameters
----------
n : int, optional
Increment to add to the internal counter of iterations
[default: 1].
"""
def close(self):
"""
Cleanup and (if leave=False) close the progressbar.
"""
def unpause(self):
"""
Restart tqdm timer from last print time.
"""
def clear(self, nomove=False):
"""
Clear current bar display
"""
def refresh(self):
"""
Force refresh the display of this bar
"""
def write(cls, s, file=sys.stdout, end="\n"):
"""
Print a message via tqdm (without overlap with bars)
"""
def set_description(self, desc=None, refresh=True):
"""
Set/modify description of the progress bar.
Parameters
----------
desc : str, optional
refresh : bool, optional
Forces refresh [default: True].
"""
def set_postfix(self, ordered_dict=None, refresh=True, **kwargs):
"""
Set/modify postfix (additional stats)
with automatic formatting based on datatype.
Parameters
----------
refresh : bool, optional
Forces refresh [default: True].
"""
- def trange(args,kwargs): """ A shortcut for tqdm(xrange(args), **kwargs).
On Python3+ range is used instead of xrange. """
- class tqdm_gui(tqdm):
""" Experimental GUI version of tqdm! """
- def tgrange(args,*kwargs):
""" Experimental GUI version of trange! """
- class tqdm_notebook(tqdm):
""" Experimental IPython/Jupyter Notebook widget using tqdm! """
- def tnrange(args,*kwargs):
""" Experimental IPython/Jupyter Notebook widget using tqdm! """
- See the examples folder;
- import the module and run
help()
, or - consult the wiki.
- this has an excellent article on how to make a great progressbar.
Custom information can be displayed and updated dynamically on tqdm
bars with the desc
and postfix
arguments:
from tqdm import trange
from random import random, randint
from time import sleep
with trange(100) as t:
for i in t:
# Description will be displayed on the left
t.set_description('GEN %i' % i)
# Postfix will be displayed on the right,
# formatted automatically based on argument's datatype
t.set_postfix(loss=random(), gen=randint(1,999), str='h',
lst=[1, 2])
sleep(0.1)
with tqdm(total=10, bar_format="{postfix[0]} {postfix[1][value]:>8.2g}",
postfix=["Batch", dict(value=0)]) as t:
for i in range(10):
sleep(0.1)
t.postfix[1]["value"] = i / 2
t.update()
Points to remember when using {postfix[...]}
in the bar_format
string:
postfix
also needs to be passed as an initial argument in a compatible format, andpostfix
will be auto-converted to a string if it is adict
-like object. To prevent this behaviour, insert an extra item into the dictionary where the key is not a string.
tqdm
supports nested progress bars. Here's an example:
On Windows colorama will be used if available to keep nested bars on their respective lines.
For manual control over positioning (e.g. for multi-threaded use), you may specify position=n
where n=0
for the outermost bar, n=1
for the next, and so on:
from time import sleep
from tqdm import trange, tqdm
from multiprocessing import Pool, freeze_support, RLock
L = list(range(9))
def progresser(n):
interval = 0.001 / (n + 2)
total = 5000
text = "#{}, est. {:<04.2}s".format(n, interval * total)
for i in trange(total, desc=text, position=n):
sleep(interval)
if __name__ == '__main__':
freeze_support() # for Windows support
p = Pool(len(L),
# again, for Windows support
initializer=tqdm.set_lock, initargs=(RLock(),))
p.map(progresser, L)
print("\n" * (len(L) - 2))
tqdm
can easily support callbacks/hooks and manual updates. Here's an example with urllib
:
urllib.urlretrieve documentation
[...]
If present, the hook function will be called once
on establishment of the network connection and once after each block read
thereafter. The hook will be passed three arguments; a count of blocks
transferred so far, a block size in bytes, and the total size of the file.
[...]
import urllib, os
from tqdm import tqdm
class TqdmUpTo(tqdm):
"""Provides `update_to(n)` which uses `tqdm.update(delta_n)`."""
def update_to(self, b=1, bsize=1, tsize=None):
"""
b : int, optional
Number of blocks transferred so far [default: 1].
bsize : int, optional
Size of each block (in tqdm units) [default: 1].
tsize : int, optional
Total size (in tqdm units). If [default: None] remains unchanged.
"""
if tsize is not None:
self.total = tsize
self.update(b * bsize - self.n) # will also set self.n = b * bsize
eg_link = "https://caspersci.uk.to/matryoshka.zip"
with TqdmUpTo(unit='B', unit_scale=True, miniters=1,
desc=eg_link.split('/')[-1]) as t: # all optional kwargs
urllib.urlretrieve(eg_link, filename=os.devnull,
reporthook=t.update_to, data=None)
Inspired by twine#242. Functional alternative in examples/tqdm_wget.py.
It is recommend to use miniters=1
whenever there is potentially large differences in iteration speed (e.g. downloading a file over a patchy connection).
Due to popular demand we've added support for pandas
-- here's an example for DataFrame.progress_apply
and DataFrameGroupBy.progress_apply
:
import pandas as pd
import numpy as np
from tqdm import tqdm
df = pd.DataFrame(np.random.randint(0, 100, (100000, 6)))
# Register `pandas.progress_apply` and `pandas.Series.map_apply` with `tqdm`
# (can use `tqdm_gui`, `tqdm_notebook`, optional kwargs, etc.)
tqdm.pandas(desc="my bar!")
# Now you can use `progress_apply` instead of `apply`
# and `progress_map` instead of `map`
df.progress_apply(lambda x: x**2)
# can also groupby:
# df.groupby(0).progress_apply(lambda x: x**2)
In case you're interested in how this works (and how to modify it for your own callbacks), see the examples folder or import the module and run help()
.
IPython/Jupyter is supported via the tqdm_notebook
submodule:
In addition to tqdm
features, the submodule provides a native Jupyter widget (compatible with IPython v1-v4 and Jupyter), fully working nested bars and color hints (blue: normal, green: completed, red: error/interrupt, light blue: no ETA); as demonstrated below.
Since tqdm
uses a simple printing mechanism to display progress bars, you should not write any message in the terminal using print()
while a progressbar is open.
To write messages in the terminal without any collision with tqdm
bar display, a .write()
method is provided:
By default, this will print to standard output sys.stdout
. but you can specify any file-like object using the file
argument. For example, this can be used to redirect the messages writing to a log file or class.
If using a library that can print messages to the console, editing the library by replacing print()
with tqdm.write()
may not be desirable. In that case, redirecting sys.stdout
to tqdm.write()
is an option.
To redirect sys.stdout
, create a file-like class that will write any input string to tqdm.write()
, and supply the arguments file=sys.stdout, dynamic_ncols=True
.
A reusable canonical example is given below:
from time import sleep
import contextlib
import sys
from tqdm import tqdm
class DummyTqdmFile(object):
"""Dummy file-like that will write to tqdm"""
file = None
def __init__(self, file):
self.file = file
def write(self, x):
# Avoid print() second call (useless \n)
if len(x.rstrip()) > 0:
tqdm.write(x, file=self.file)
def flush(self):
return getattr(self.file, "flush", lambda: None)()
@contextlib.contextmanager
def std_out_err_redirect_tqdm():
orig_out_err = sys.stdout, sys.stderr
try:
sys.stdout, sys.stderr = map(DummyTqdmFile, orig_out_err)
yield orig_out_err[0]
# Relay exceptions
except Exception as exc:
raise exc
# Always restore sys.stdout/err if necessary
finally:
sys.stdout, sys.stderr = orig_out_err
def some_fun(i):
print("Fee, fi, fo,".split()[i])
# Redirect stdout to tqdm.write() (don't forget the `as save_stdout`)
with std_out_err_redirect_tqdm() as orig_stdout:
# tqdm needs the original stdout
# and dynamic_ncols=True to autodetect console width
for i in tqdm(range(3), file=orig_stdout, dynamic_ncols=True):
sleep(.5)
some_fun(i)
# After the `with`, printing is restored
print("Done!")
tqdm
implements a few tricks to to increase efficiency and reduce overhead.
- Avoid unnecessary frequent bar refreshing:
mininterval
defines how long to wait between each refresh.tqdm
always gets updated in the background, but it will diplay only everymininterval
. - Reduce number of calls to check system clock/time.
mininterval
is more intuitive to configure thanminiters
. A clever adjustment systemdynamic_miniters
will automatically adjustminiters
to the amount of iterations that fit into timemininterval
. Essentially,tqdm
will check if it's time to print without actually checking time. This behaviour can be still be bypassed by manually settingminiters
.
However, consider a case with a combination of fast and slow iterations. After a few fast iterations, dynamic_miniters
will set miniters
to a large number. When iteration rate subsequently slows, miniters
will remain large and thus reduce display update frequency. To address this:
maxinterval
defines the maximum time between display refreshes. A concurrent monitoring thread checks for overdue updates and forces one where necessary.
The monitoring thread should not have a noticeable overhead, and guarantees updates at least every 10 seconds by default. This value can be directly changed by setting the monitor_interval
of any tqdm
instance (i.e. t = tqdm.tqdm(...); t.monitor_interval = 2
). The monitor thread may be disabled application-wide by setting tqdm.tqdm.monitor_interval = 0
before instantiatiation of any tqdm
bar.
All source code is hosted on GitHub. Contributions are welcome.
See the CONTRIBUTING file for more information.
The main developers, ranked by surviving lines of code, are:
- Casper da Costa-Luis (casperdcl, ~2/3,
)
- Stephen Larroque (lrq3000, ~1/3)
- Guangshuo Chen (chengs, ~1%)
- Hadrien Mary (hadim, ~1%)
- Noam Yorav-Raphael (noamraph, ~1%, original author)
- Mikhail Korobov (kmike, ~1%)