tensorboardY 0.1.7

Interact with ML models in the browser so that we can better understand their strengths and weaknesses on real world data

tensorboardY

The goal of this library is to accelerate ML research by making it as easy as possible to demo ML algorithms in the browser. The easier it is to interact with ML models, the faster we can determine their current limitations.

This library seeks to automate the creation of cool ML demo websites like

• Megadepth
• Talk to transformer
• GLTR
• Wayve demo
• FUNIT

Install

pip install tensorboardY

Examples

• python examples/simple_example.py

import tensorboardY as ty
import matplotlib.pyplot as plt

def forward(x, title):
    plt.imshow(x)
    plt.title(title)
    return plt.gcf()

inputs = [ty.Image(var='x', exs=['imgs/curve.jpg']),
          ty.Text(var='title', exs=["EXAMPLE"])]

ty.show(forward, inputs)

• python examples/full_example.py

import tensorboardY as ty
import os

def forward(z):
    return z

inputs = [ty.Widget("z", name="Choose your input",
                    camera=True,
                    image_upload=True,
                    image_list=['imgs/curve.jpg'], image_names=['Curve example'],
                    text_input=True,
                    text_list=['This is an example text!'], text_names=['Random'],
                    option_list=["This is an example option!"], option_names=['Resnet50'],
                    boolean=True,
                    slider=(5, 20, 0.5), slider_default=10.3)]

ty.show(forward, inputs)