mlvt

mlvt copied to clipboard

Machine Learning Visualization Tools

A single Python file with some tools for visualizing machine learning in the terminal.

https://user-images.githubusercontent.com/4842908/178077497-f5c202bf-5cba-4423-a5c5-9b08ad4b5d6f.mov

This demo is composed of three ideas, which are explained below. Here's how to get started:

pip install git+https://github.com/bwasti/mlvt.git

or just copy the mlvt.py file!

Then, grab the test.py file and run it on your machine:

curl https://raw.githubusercontent.com/bwasti/mlvt/main/test.py > test_mlvt.py
python mlvt.py

Pick and choose parts of that file for your own use. :) (Remember to use rp.flush()!)

mlvt.Reprint

Reprint helps with in-line animations. It works by keeping track of how much it printed so far and reprinting it when flush() is called.

You can use the with statement to hijack print statements and auto_flush=True to avoid calling flush() in a loop, like so:

print("loading!")
with mlvt.Reprint(auto_flush=True) as rp:
  for i in range(100):
    print(f"{i+1}%") # Reprint detects the loop and overwrites in-place
    time.sleep(0.02)
print("done!")

or, if you'd prefer to avoid contexts, loop-detection and hijacked builtins

print("loading!")
rp = mlvt.Reprint()
for i in range(100):
  rp.print(f"{i+1}%")
  rp.flush()
  time.sleep(0.02)
print("done!")

mlvt.horiz_concat

horiz_concat concatenates multi-line strings horizontally, accounting for padding and ANSI escape sequences (for color text).

a = """
{ hello! }
          \_    
"""
b = """
 ___
|. .|
| ^ |
| o |
"""
print(mlvt.horiz_concat(a, b, padding=2))

yields

               ___
{ hello! }    |. .|
          \_  | ^ |
              | o |
              

plotille wrappers

Finally, there are a couple of small plotille wrappers that decouple updating charts and printing them. That library is great on its own, so I encourage you to check it out!

import mlvt
import numpy as np

# all charts take in width, height, color
hist = mlvt.Histogram(32, 8, color="bright_blue")
hist.update(np.random.randn(100))
print(hist)

gives us

 (Counts)  ^
8.80000000 |
7.70000000 | ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢠⡄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
6.60000000 | ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢸⡇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
5.50000000 | ⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢸⡇⣶⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
4.40000000 | ⠀⠀⠀⠀⠀⠀⠀⠀⢰⣶⣶⠀⠀⢸⡇⣿⠀⢰⣶⡆⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
3.30000000 | ⠀⠀⠀⠀⠀⠀⠀⣿⢸⣿⣿⣿⣿⢸⣿⣿⣿⢸⣿⣿⣿⣿⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
2.20000000 | ⠀⠀⠀⠀⠀⠀⠀⣿⣿⣿⣿⣿⣿⢸⣿⣿⣿⣿⣿⣿⣿⣿⡇⢸⣿⠀⠀⠀⢸⡇⠀⠀
1.10000000 | ⠀⠀⢀⣀⡀⣿⣀⣿⣿⣿⣿⣿⣿⣸⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠀⠀⣇⣸⡇⠀⠀
         0 | ⠀⠀⢸⣿⡇⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠀⠀⣿⣿⡇⠀⠀
-----------|-|---------|---------|---------|-> (X)
           | -2.124059 -0.741902 0.6402548 2.0224115