Skip to content

tqdm.md

Latest commit

 

History

History
83 lines (68 loc) · 4.23 KB

tqdm.md

File metadata and controls

83 lines (68 loc) · 4.23 KB

tqdm — progress bars

Version: 4.67.3
Type: Pure Python
SPM target: Tqdm
Auto-included by: huggingface_hub, transformers, datasets, manim, pip
Total Python modules: 31

Wrap any iterable with tqdm(...) to get a live progress bar with ETA, rate, postfix stats. Multiple flavors (stdout, GUI, Jupyter, rich, async). On iOS the stdout flavor renders as in-place updates in the in-app shell.

Modules

Top-level

Module What it does
tqdm.__init__ Re-exports tqdm, trange, tqdm_pandas, tqdm_gui, tgrange, TMonitor, TqdmSynchronisationWarning, __version__, main
tqdm.std The reference tqdm class (used by from tqdm import tqdm). Also Bar, EMA, TqdmDefaultWriteLock, TRLock, all warning types (TqdmTypeError, TqdmKeyError, TqdmWarning, TqdmDeprecationWarning, TqdmMonitorWarning, TqdmExperimentalWarning)
tqdm.auto Auto-detect best flavor — uses tqdm.notebook in Jupyter, tqdm.std elsewhere. The recommended import: from tqdm.auto import tqdm
tqdm.autonotebook Same as auto but errors-loud if not in a notebook
tqdm.asyncio tqdm.asyncio.tqdm — async iterator support: async for x in tqdm(async_iter)
tqdm.gui Tkinter / matplotlib GUI bars — no GUI on iOS, falls back to stdout
tqdm.notebook Jupyter widget bars (IProgress HTML widget)
tqdm.tk Tk-based bar window
tqdm.rich Rich-rendered bar — uses rich.progress.Progress
tqdm.dask tqdm.dask.TqdmCallback — Dask scheduler progress
tqdm.keras tqdm.keras.TqdmCallback — Keras training callback
tqdm.cli tqdm command-line entry — pipe stdin through a progress bar
tqdm.version __version__ lookup (via importlib.metadata)
tqdm.utils disp_len, disp_trim, IS_WIN, IS_NIX, color helpers
tqdm._utils Deprecated alias for tqdm.utils (re-exported for back-compat)
tqdm._monitor TMonitor thread that warns about hung iteration
tqdm._main Deprecated alias for tqdm.cli
tqdm._tqdm / _tqdm_gui / _tqdm_notebook / _tqdm_pandas Deprecated aliases for tqdm.std / gui / notebook / pandas integration

tqdm.contrib — extras

Submodule What it does
contrib.__init__ tenumerate, tzip, tmap — wrapped enumerate/zip/map
contrib.bells tqdm with all extras enabled (rich + auto + colour)
contrib.concurrent process_map, thread_map — concurrent.futures + progress bar
contrib.discord Report progress to a Discord channel via webhook
contrib.slack Report progress to Slack via webhook
contrib.telegram Report progress to a Telegram chat
contrib.logging tqdm_logging_redirect ctx manager — route logging writes through tqdm.write so log lines don't break the bar
contrib.itertools product — itertools.product with progress
contrib.utils_worker MonoWorker — worker that prints each result on its own line

iOS notes

  • stdout bar works. Carriage-return based redraw renders cleanly in the in-app shell.
  • No Jupyter widgets. tqdm.notebook imports ipywidgets which isn't shipped. Use tqdm.auto — it falls back to tqdm.std.
  • No GUI. tqdm.gui / tqdm.tk need Tk/matplotlib displays — non-functional on iOS.
  • huggingface_hub and transformers log via tqdm. Model download bars work out of the box.
  • TMonitor thread: disable via tqdm.set_lock(None) or TQDM_DISABLE_MONITOR=1 if you see "iteration stuck" false positives on iOS (slow filesystem operations can trip the 60-second default).

Example

from tqdm.auto import tqdm, trange
from tqdm.contrib.concurrent import thread_map
import time

# Wrap any iterable
for path in tqdm(file_list, desc="Hashing", unit="file"):
    hash_file(path)

# Manual / unknown total
pbar = tqdm(total=None, desc="Streaming", unit="B", unit_scale=True)
while chunk := stream.read(8192):
    pbar.update(len(chunk))
pbar.close()

# Parallel + progress in one line
results = thread_map(download, urls, max_workers=4, desc="Downloading")

# Postfix stats
pbar = trange(100, desc="Training")
for i in pbar:
    loss = train_step()
    pbar.set_postfix(loss=f"{loss:.4f}", lr=1e-3)