Waffle is a framework that lets you use lots of different deep learning tools through just one interface. When it comes to MLOps (machine learning operations), you need to be able to keep up with all the new ideas in deep learning as quickly as possible. But it's hard to do that if you have to write all the code yourself. That's why we started a project to bring together different tools into one framework.

Experience the power of multiple deep learning frameworks at your fingertips with Waffle's seamless integration, unlocking limitless possibilities for your machine learning projects.

Prerequisites

We've tested Waffle on the following environments:

OS	Python	PyTorch	Device	Backend
Ubuntu 20.04	3.9, 3.10	1.13.1	CPU, GPU	All
Windows	3.9, 3.10	1.13.1	CPU, GPU	All
Ubuntu 20.04	3.9	1.13.1	Multi GPU	Ultralytics

We recommend using above environments for the best experience.

Installation

Install pytorch and torchvision
- PyTorch and TorchVision (We recommend using 1.13.1)
Install Waffle Hub
- pip install -U waffle-hub

Example Usage

We provide both python module and CLI for Waffle Hub.

Following examples do the exact same thing.

Python Module

from waffle_hub.dataset import Dataset
dataset = Dataset.sample(
  name = "mnist_classification",
  task = "classification",
)
dataset.split(
  train_ratio = 0.8,
  val_ratio = 0.1,
  test_ratio = 0.1
)
export_dir = dataset.export("YOLO")

from waffle_hub.hub import Hub
hub = Hub.new(
  name = "my_classifier",
  task = "classification",
  model_type = "yolov8",
  model_size = "n",
  categories = dataset.get_category_names(),
)
hub.train(
  dataset = dataset,
  epochs = 30,
  batch_size = 64,
  image_size=64,
  device="cpu"
)
hub.inference(
  source=export_dir,
  draw=True,
  device="cpu"
)

CLI

wd sample --name mnist_classification --task classification
wd split --name mnist_classification --train-ratio 0.8 --val-ratio 0.1 --test-ratio 0.1
wd export --name mnist_classification --data-type YOLO

wh new --name my_classifier --task classification --model-type yolov8 --model-size n --categories [1,2]
wh train --name my_classifier --dataset mnist_classification --epochs 30 --batch-size 64 --image-size 64 --device cpu
wh inference --name my_classifier --source datasets/mnist_classification/exports/YOLO --draw --device cpu

See our documentation for more information!

Metadata

Release files for waffle-hub 0.3.1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for waffle-hub 0.3.1
File	Size	Uploaded
waffle_hub-0.3.1.tar.gz	97.0 kB	Jan 22, 2024	Details

Built distribution (wheel)

Table of built distributions (wheels) for waffle-hub 0.3.1
File	Interpreter	ABI	Platform	Reset
waffle_hub-0.3.1-py3-none-any.whl 109.0 kB Jan 22, 2024	Python 3	none	any	Details

Total release size: 206.0 kB

Release files / waffle_hub-0.3.1.tar.gz

Download URL	waffle_hub-0.3.1.tar.gz
Size	97.0 kB
Tags	Source
SHA-256 checksum How to use checksums	`ae837f987b83abd912a2b9a8b680bc3cab24088f03bdc9d665e8d57d3404dee7`
BLAKE2b-256 checksum How to use checksums	`0cf9276a5f342ffdb636ca9a831ad0d7fe0c6c40059b3fc41efb5846bad27f0e`
Upload date	Jan 22, 2024
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/4.0.2 CPython/3.9.18`

Release files / waffle_hub-0.3.1-py3-none-any.whl

Download URL	waffle_hub-0.3.1-py3-none-any.whl
Size	109.0 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`6ec8b312c6d1df389409faf11d474d4d65aa597c5516110f174625fa63ff8af8`
BLAKE2b-256 checksum How to use checksums	`185a1299b89952a8397e20c4d777829e85c34ce06139078065e2fd38d2d81b30`
Upload date	Jan 22, 2024
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/4.0.2 CPython/3.9.18`