tqdm_progress_bar

TQDMProgressBar(
    metrics_separator: str = ' - ',
    overall_bar_format: str = '{l_bar}{bar} {n_fmt}/{total_fmt} ETA: {remaining}s,  {rate_fmt}{postfix}',
    epoch_bar_format: str = '{n_fmt}/{total_fmt}{bar} ETA: {remaining}s - {desc}',
    metrics_format: str = '{name}: {value:0.4f}',
    update_per_second: int = 10,
    leave_epoch_progress: bool = True,
    leave_overall_progress: bool = True,
    show_epoch_progress: bool = True,
    show_overall_progress: bool = True,
)

TQDM Progress Bar callback.

Example:

import time
import helia_edge as helia

# Create a TQDM progress bar
pb_callback = helia.callbacks.TQDMProgressBar(
    overall_bar_format="{l_bar}{bar} {n_fmt}/{total_fmt},  {rate_fmt}{postfix}",
    epoch_bar_format="{n_fmt}/{total_fmt}{bar} - {desc}",
    metrics_format="{name}: {value:0.4f}",
    update_per_second=10,
    leave_epoch_progress=True,
)

# Simulate training
epochs = 10
steps = 5
pb_callback.set_params(dict(epochs=epochs, steps=steps))
pb_callback.on_train_begin()
loss = 1.0
accuracy = 0.0
for epoch in range(epochs):
    pb_callback.on_epoch_begin(epoch)
    for step in range(steps):
        loss -= epoch * step / (epochs * steps)
        accuracy += epoch * step / (epochs * steps)
        pb_callback.on_batch_end(step, {"loss": loss, "accuracy": accuracy})
        time.sleep(0.1)
    pb_callback.on_epoch_end(epoch, {"loss": loss, "accuracy": accuracy})
pb_callback.on_train_end()

Parameters

Parameters of TQDMProgressBar
Name	Type	Default	Description
`metrics_separator`	`str`	' - '	Custom separator between metrics. Defaults to ' - '.
`overall_bar_format`	`str`	'{l_bar}{bar} {n_fmt}/{total_fmt} ETA: {remaining}s, {rate_fmt}{postfix}'	Custom bar format for overall (outer) progress bar, see https://github.com/tqdm/tqdm#parameters for more detail.
`epoch_bar_format`	`str`	'{n_fmt}/{total_fmt}{bar} ETA: {remaining}s - {desc}'	Custom bar format for epoch (inner) progress bar, see https://github.com/tqdm/tqdm#parameters for more detail.
`update_per_second`	`int`	10	Maximum number of updates in the epochs bar per second, this is to prevent small batches from slowing down training. Defaults to 10.
`metrics_format`	`str`	'{name}: {value:0.4f}'	Custom format for how metrics are formatted. See https://github.com/tqdm/tqdm#parameters for more detail.
`leave_epoch_progress`	`bool`	True	`True` to leave epoch progress bars.
`leave_overall_progress`	`bool`	True	`True` to leave overall progress bar.
`show_epoch_progress`	`bool`	True	`False` to hide epoch progress bars.
`show_overall_progress`	`bool`	True	`False` to hide overall progress bar.

metrics_separator = metrics_separator

overall_bar_format = overall_bar_format

epoch_bar_format = epoch_bar_format

leave_epoch_progress = leave_epoch_progress

leave_overall_progress = leave_overall_progress

show_epoch_progress = show_epoch_progress

show_overall_progress = show_overall_progress

metrics_format = metrics_format

update_interval = 1 / update_per_second

last_update_time = time.time()

overall_progress_tqdm = None

epoch_progress_tqdm = None

is_training = False

num_epochs = None

logs = None

on_train_begin(logs=None)

on_train_end(logs={})

on_test_begin(logs={})

on_test_end(logs={})

on_epoch_begin(epoch, logs={})

on_epoch_end(epoch, logs={})

on_test_batch_end(batch, logs={})

on_batch_end(batch, logs={})

format_metrics(logs: dict = {}, factor=1) -> str

Format metrics in logs into a string.

Parameters

Parameters of format_metrics
Name	Type	Default	Description
`logs`	`dict`	{}	dictionary of metrics and their values. Defaults to empty dictionary.
`factor`	`int`	1	The factor we want to divide the metrics in logs by, useful when we are computing the logs after each batch. Defaults to 1.

Returns

Returns of format_metrics
Value	Type	Description
`metrics_string`	`str`	a string displaying metrics using the given
	`str`	formators passed in through the constructor.

get_config()