toolman 0.1.18

Python utility tools for research

See the source for this project on our GitHub page

Install

pip install toolman

Modules

mist_utils.py

miscellaneous utility functions including data I/O and processing

1. Read/write different formats of files in one function:

import toolman as tm
data = tm.misc_utils.load_file(file_name)
tm.misc_utils.save_file(file_name, data)

Currently support extensions including: .npy, .pkl, .txt, .csv, .json and commonly used image formats.

2. Argument parser, parse nested argument list:

import sys
import argparse
import toolman as tm
parser = argparse.ArgumentParser()
args, extras = parser.parse_known_args(sys.argv[1:])
cfg_dict = tm.misc_utils.parse_args(extras)

vis_utils.py

Matplotlib utility functions for visualization

1. Display images in side by side with axis linked

import toolman as tm
fig1 = tm.misc_utils.load_file(img_name_1)
fig2 = tm.misc_utils.load_file(img_name_1)
tm.vis_utils.compare_figures([fig1, fig2], (1, 2), fig_size=(12, 5))

2. Display barplots

import toolman as tm
data = np.random.random((3, 4))
labels = ['group 1', 'group 2', 'group 3']
xticks = ['cluster 1', 'cluster 2', 'cluster 3', 'cluster 4']
tm.vis_utils.compare_bars(data, labels, xticks)

img_utils

image specific utility functions

pytorch_utils

pytorch specific utility functions

process_block

A processing unit that do certain operations only if it has never done before. This is helpful avoid duplicate executing time consuming jobs.

import toolman as tm
def foo(cnt_len):
    cnt = 0
    for i in range(cnt_len):
        cnt += 1
    return cnt

pb = tm.process_block.ProcessBlock(foo, file_dir)
pb.run(force_run=False, cnt_len=100)