Kindle - PyTorch no-code model builder

`Documentation`

Kindle is an easy model build package for PyTorch. Building a deep learning model became so simple that almost all model can be made by copy and paste from other existing model codes. So why code? when we can simply build a model with yaml markup file.

Kindle builds a model with no code but yaml file which its method is inspired from YOLOv5.

Installation
AutoML with Kindle
Usage
Supported modules
Custom module support
- Custom module with yaml
- Custom module from source

Installation

Install with pip

PyTorch is required prior to install. Please visit PyTorch installation guide to install.

You can install kindle by pip.

$ pip install kindle

Install from source

Please visit Install from source wiki page

For contributors

Please visit For contributors wiki page

Usage

Build a model

Make model yaml file

Example model https://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html

input_size: [32, 32]
input_channel: 3

depth_multiple: 1.0
width_multiple: 1.0

backbone:
    # [from, repeat, module, args]
    [
        [-1, 1, Conv, [6, 5, 1, 0]],
        [-1, 1, MaxPool, [2]],
        [-1, 1, Conv, [16, 5, 1, 0]],
        [-1, 1, MaxPool, [2]],
        [-1, 1, Flatten, []],
        [-1, 1, Linear, [120, ReLU]],
        [-1, 1, Linear, [84, ReLU]],
        [-1, 1, Linear, [10]]
    ]

Build the model with kindle

from kindle import Model

model = Model("model.yaml"), verbose=True)

idx |       from |   n |     params |          module |            arguments |                       in shape |       out shape |
---------------------------------------------------------------------------------------------------------------------------------
  0 |         -1 |   1 |        616 |            Conv |         [6, 5, 1, 0] |                    [3, 32, 32] |     [8, 32, 32] |
  1 |         -1 |   1 |          0 |         MaxPool |                  [2] |                      [8 32 32] |     [8, 16, 16] |
  2 |         -1 |   1 |      3,232 |            Conv |        [16, 5, 1, 0] |                      [8 16 16] |    [16, 16, 16] |
  3 |         -1 |   1 |          0 |         MaxPool |                  [2] |                     [16 16 16] |      [16, 8, 8] |
  4 |         -1 |   1 |          0 |         Flatten |                   [] |                       [16 8 8] |          [1024] |
  5 |         -1 |   1 |    123,000 |          Linear |        [120, 'ReLU'] |                         [1024] |           [120] |
  6 |         -1 |   1 |     10,164 |          Linear |         [84, 'ReLU'] |                          [120] |            [84] |
  7 |         -1 |   1 |        850 |          Linear |                 [10] |                           [84] |            [10] |
Model Summary: 21 layers, 137,862 parameters, 137,862 gradients

AutoML with Kindle

Kindle offers the easiest way to build your own deep learning architecture. Beyond building a model, AutoML became easier with Kindle and Optuna or other optimization frameworks.
For further information, please refer to here

Supported modules

Detailed documents can be found here

Module	Components	Arguments
Conv	Conv -> BatchNorm -> Activation	[channel, kernel size, stride, padding, activation]
DWConv	DWConv -> BatchNorm -> Activation	[channel, kernel_size, stride, padding, activation]
Bottleneck	Expansion ConvBNAct -> ConvBNAct	[channel, shortcut, groups, expansion, activation]
AvgPool	Average pooling	[kernel_size, stride, padding]
MaxPool	Max pooling	[kernel_size, stride, padding]
GlobalAvgPool	Global Average Pooling	[]
Flatten	Flatten	[]
Concat	Concatenation	[dimension]
Linear	Linear	[channel, activation]
Add	Add	[]

Custom module support

Custom module with yaml

You can make your own custom module with yaml file.

1. custom_module.yaml

args: [96, 32]

module:
    # [from, repeat, module, args]
    [
        [-1, 1, Conv, [arg0, 1, 1]],
        [0, 1, Conv, [arg1, 3, 1]],
        [0, 1, Conv, [arg1, 5, 1]],
        [0, 1, Conv, [arg1, 7, 1]],
        [[1, 2, 3], 1, Concat, [1]],
        [[0, 4], 1, Add, []],
    ]

Arguments of yaml module can be defined as arg0, arg1 ...

2. model_with_custom_module.yaml

input_size: [32, 32]
input_channel: 3

depth_multiple: 1.0
width_multiple: 1.0

backbone:
    [
        [-1, 1, Conv, [6, 5, 1, 0]],
        [-1, 1, MaxPool, [2]],
        [-1, 1, YamlModule, ["custom_module.yaml", 48, 16]],
        [-1, 1, MaxPool, [2]],
        [-1, 1, Flatten, []],
        [-1, 1, Linear, [120, ReLU]],
        [-1, 1, Linear, [84, ReLU]],
        [-1, 1, Linear, [10]]
    ]

Note that argument of yaml module can be provided.

3. Build model

from kindle import Model

model = Model("model_with_custom_module.yaml"), verbose=True)

idx |       from |   n |     params |          module |            arguments |                       in shape |       out shape |
---------------------------------------------------------------------------------------------------------------------------------
  0 |         -1 |   1 |        616 |            Conv |         [6, 5, 1, 0] |                    [3, 32, 32] |     [8, 32, 32] |
  1 |         -1 |   1 |          0 |         MaxPool |                  [2] |                      [8 32 32] |     [8, 16, 16] |
  2 |         -1 |   1 |     10,832 |      YamlModule |    ['custom_module'] |                      [8 16 16] |    [24, 16, 16] |
  3 |         -1 |   1 |          0 |         MaxPool |                  [2] |                     [24 16 16] |      [24, 8, 8] |
  4 |         -1 |   1 |          0 |         Flatten |                   [] |                       [24 8 8] |          [1536] |
  5 |         -1 |   1 |    184,440 |          Linear |        [120, 'ReLU'] |                         [1536] |           [120] |
  6 |         -1 |   1 |     10,164 |          Linear |         [84, 'ReLU'] |                          [120] |            [84] |
  7 |         -1 |   1 |        850 |          Linear |                 [10] |                           [84] |            [10] |
Model Summary: 36 layers, 206,902 parameters, 206,902 gradients

Custom module from source

You can make your own custom module from the source.

1. custom_module_model.yaml

input_size: [32, 32]
input_channel: 3

depth_multiple: 1.0
width_multiple: 1.0

custom_module_paths: ["tests.test_custom_module"]  # Paths to the custom modules of the source

backbone:
    # [from, repeat, module, args]
    [
        [-1, 1, MyConv, [6, 5, 3]],
        [-1, 1, MaxPool, [2]],
        [-1, 1, MyConv, [16, 3, 5, SiLU]],
        [-1, 1, MaxPool, [2]],
        [-1, 1, Flatten, []],
        [-1, 1, Linear, [120, ReLU]],
        [-1, 1, Linear, [84, ReLU]],
        [-1, 1, Linear, [10]]
    ]

2. Write PyTorch module and ModuleGenerator

tests/test_custom_module.py

from typing import List, Union

import numpy as np
import torch
from torch import nn

from kindle.generator import GeneratorAbstract
from kindle.torch_utils import Activation, autopad


class MyConv(nn.Module):
    def __init__(
        self,
        in_channels: int,
        out_channels: int,
        kernel_size: int,
        n: int,
        activation: Union[str, None] = "ReLU",
    ) -> None:
        super().__init__()
        convs = []
        for i in range(n):
            convs.append(
                nn.Conv2d(
                    in_channels,
                    in_channels if (i + 1) != n else out_channels,
                    kernel_size,
                    padding=autopad(kernel_size),
                    bias=False,
                )
            )

        self.convs = nn.Sequential(*convs)
        self.batch_norm = nn.BatchNorm2d(out_channels)
        self.activation = Activation(activation)()

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        return self.activation(self.batch_norm(self.convs(x)))


class MyConvGenerator(GeneratorAbstract):
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)

    @property
    def out_channel(self) -> int:
        return self._get_divisible_channel(self.args[0] * self.width_multiply)

    @property
    def in_channel(self) -> int:
        if isinstance(self.from_idx, list):
            raise Exception("from_idx can not be a list.")
        return self.in_channels[self.from_idx]

    @torch.no_grad()
    def compute_out_shape(self, size: np.ndarray, repeat: int = 1) -> List[int]:
        module = self(repeat=repeat)
        module.eval()
        module_out = module(torch.zeros([1, *list(size)]))
        return list(module_out.shape[-3:])

    def __call__(self, repeat: int = 1) -> nn.Module:
        args = [self.in_channel, self.out_channel, *self.args[1:]]
        if repeat > 1:
            module = [MyConv(*args) for _ in range(repeat)]
        else:
            module = MyConv(*args)

        return self._get_module(module)

3. Build a model

from kindle import Model

model = Model("custom_module_model.yaml"), verbose=True)

idx |       from |   n |     params |          module |            arguments |                       in shape |       out shape |
---------------------------------------------------------------------------------------------------------------------------------
  0 |         -1 |   1 |      1,066 |          MyConv |            [6, 5, 3] |                    [3, 32, 32] |     [8, 32, 32] |
  1 |         -1 |   1 |          0 |         MaxPool |                  [2] |                      [8 32 32] |     [8, 16, 16] |
  2 |         -1 |   1 |      3,488 |          MyConv |   [16, 3, 5, 'SiLU'] |                      [8 16 16] |    [16, 16, 16] |
  3 |         -1 |   1 |          0 |         MaxPool |                  [2] |                     [16 16 16] |      [16, 8, 8] |
  4 |         -1 |   1 |          0 |         Flatten |                   [] |                       [16 8 8] |          [1024] |
  5 |         -1 |   1 |    123,000 |          Linear |        [120, 'ReLU'] |                         [1024] |           [120] |
  6 |         -1 |   1 |     10,164 |          Linear |         [84, 'ReLU'] |                          [120] |            [84] |
  7 |         -1 |   1 |        850 |          Linear |                 [10] |                           [84] |            [10] |
Model Summary: 29 layers, 138,568 parameters, 138,568 gradients

Planned features

~~Custom module support~~
~~Custom module with yaml support~~
Use pre-trained model
More modules!

Kindle is an easy model build package for PyTorch.

Related tags

Overview

Kindle - PyTorch no-code model builder

Contents

Installation

Install with pip

Install from source

For contributors

Usage

Build a model

AutoML with Kindle

Supported modules

Custom module support

Custom module with yaml

Custom module from source

Planned features

Owner

Jongkuk Lim

"Reinforcement Learning for Bandit Neural Machine Translation with Simulated Human Feedback"

Certified Patch Robustness via Smoothed Vision Transformers

A general 3D Object Detection codebase in PyTorch.

The implementation code for "DAGAN: Deep De-Aliasing Generative Adversarial Networks for Fast Compressed Sensing MRI Reconstruction"

The LaTeX and Python code for generating the paper, experiments' results and visualizations reported in each paper is available (whenever possible) in the paper's directory

Look Closer: Bridging Egocentric and Third-Person Views with Transformers for Robotic Manipulation

Disentangled Lifespan Face Synthesis

Multi Agent Path Finding Algorithms

No-Reference Image Quality Assessment via Transformers, Relative Ranking, and Self-Consistency

HairCLIP: Design Your Hair by Text and Reference Image

Optimizing Value-at-Risk and Conditional Value-at-Risk of Black Box Functions with Lacing Values (LV)

Learning to Communicate with Deep Multi-Agent Reinforcement Learning in PyTorch

Official code for UnICORNN (ICML 2021)

Python script for performing depth completion from sparse depth and rgb images using the msg_chn_wacv20. model in Tensorflow Lite.

This is the code for our paper "Iconary: A Pictionary-Based Game for Testing Multimodal Communication with Drawings and Text"

《LightXML: Transformer with dynamic negative sampling for High-Performance Extreme Multi-label Text Classiﬁcation》(AAAI 2021) GitHub:

Forecasting for knowable future events using Bayesian informative priors (forecasting with judgmental-adjustment).

A Japanese Medical Information Extraction Toolkit

A way to store images in YAML.

DynaTune: Dynamic Tensor Program Optimization in Deep Neural Network Compilation