Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/Atharva-Phatak/torchflare
TorchFlare is a simple, beginner-friendly, and easy-to-use PyTorch Framework train your models effortlessly.
https://github.com/Atharva-Phatak/torchflare
deep-learning deep-neural-networks gans generative-adversarial-network image-classfication image-classification image-processing image-segmentation infrastructure kaggle machine-learning natural-language-processing python pytorch pytorch-model self-supervised-learning text-classification
Last synced: about 2 months ago
JSON representation
TorchFlare is a simple, beginner-friendly, and easy-to-use PyTorch Framework train your models effortlessly.
- Host: GitHub
- URL: https://github.com/Atharva-Phatak/torchflare
- Owner: Atharva-Phatak
- License: apache-2.0
- Created: 2021-04-16T14:55:28.000Z (over 3 years ago)
- Default Branch: main
- Last Pushed: 2024-07-29T20:12:31.000Z (about 2 months ago)
- Last Synced: 2024-07-30T02:24:08.301Z (about 2 months ago)
- Topics: deep-learning, deep-neural-networks, gans, generative-adversarial-network, image-classfication, image-classification, image-processing, image-segmentation, infrastructure, kaggle, machine-learning, natural-language-processing, python, pytorch, pytorch-model, self-supervised-learning, text-classification
- Language: Python
- Homepage: https://torchflare.readthedocs.io/en/latest/
- Size: 18.1 MB
- Stars: 84
- Watchers: 4
- Forks: 7
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- Contributing: .github/CONTRIBUTING.MD
- License: LICENSE
- Code of conduct: CODE_OF_CONDUCT.md
Awesome Lists containing this project
README
[](https://www.codefactor.io/repository/github/atharva-phatak/torchflare)
[](https://github.com/Atharva-Phatak/torchflare/actions/workflows/test.yml)
[](https://torchflare.readthedocs.io/en/latest/?badge=latest)
[](https://github.com/Atharva-Phatak/torchflare/actions/workflows/publish.yml)
[](https://deepsource.io/gh/Atharva-Phatak/torchflare/?ref=repository-badge)
[](https://deepsource.io/gh/Atharva-Phatak/torchflare/?ref=repository-badge)
[](https://codecov.io/gh/Atharva-Phatak/torchflare)
[](https://www.python.org/)
[](https://github.com/Atharva-Phatak/torchflare/blob/main/LICENSE)
[](https://github.com/psf/black)
# ***TorchFlare***
***TorchFlare*** is a simple, beginner-friendly and an easy-to-use PyTorch Framework to train your models with ease.
It provides an almost Keras-like experience for training
your models with all the callbacks, metrics, etc
### ***Features***
* _A high-level module for Keras-like training._
* _Flexibility to write custom training and validation loops for advanced use cases._
* _Off-the-shelf Pytorch style Datasets/Dataloaders for standard tasks such as **Image classification, Image segmentation,
Text Classification**, etc_
* _**Callbacks** for model checkpoints, early stopping, and much more!_
* _**TorchFlare** uses powerful [torchmetrics](https://github.com/PyTorchLightning/metrics) in the backend for metric computations!_
* _**Reduction** of the boiler plate code required for training your models._
* _Create **interactive UI** for model prototyping and POC_
***
Currently, **TorchFlare** supports ***CPU*** and ***GPU*** training. DDP and TPU support will be coming soon!
***
### ***Installation***
pip install torchflare
***
### ***Documentation***
The Documentation is available [here](https://torchflare.readthedocs.io/en/latest/)
***
### ***Getting Started***
The core idea around TorchFlare is the [Experiment](/torchflare/experiments/experiment.py)
class. It handles all the internal stuff like boiler plate code for training,
calling callbacks,metrics,etc. The only thing you need to focus on is creating you PyTorch Model.
Also, there are off-the-shelf pytorch style datasets/dataloaders available for standard tasks, so that you don't
have to worry about creating Pytorch Datasets/Dataloaders.
Here is an easy-to-understand example to show how Experiment class works.
``` python
import torch
import torchmetrics
import torch.nn as nn
from torchflare.experiments import Experiment, ModelConfig
import torchflare.callbacks as cbs
# Some dummy dataloaders
train_dl = SomeTrainingDataloader()
valid_dl = SomeValidationDataloader()
test_dl = SomeTestingDataloader()
```
Create a pytorch Model
``` python
class Net(nn.Module):
def __init__(self, n_classes, p_dropout):
super().__init__()
self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
self.conv2_drop = nn.Dropout2d(p=p_dropout)
self.fc1 = nn.Linear(320, 50)
self.fc2 = nn.Linear(50, n_classes)
def forward(self, x):
x = F.relu(F.max_pool2d(self.conv1(x), 2))
x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2))
x = x.view(-1, 320)
x = F.relu(self.fc1(x))
x = F.dropout(x, training=self.training)
x = self.fc2(x)
return x
```
Define callbacks and metrics
``` python
metric_list = [
torchmetrics.Accuracy(num_classes=num_classes)
]
callbacks = [
cbs.EarlyStopping(monitor="val_accuracy", mode="max"),
cbs.ModelCheckpoint(monitor="val_accuracy"),
cbs.ReduceLROnPlateau(mode="max", patience=2),
]
```
Define Model Configuration
``` python
#Defining Model Config for experiment.
config = ModelConfig(
nn_module=Net,
module_params={"n_classes": 10, "p_dropout": 0.3},
optimizer="Adam",
optimizer_params={"lr": 3e-4},
criterion="cross_entropy",
)
```
Define your experiment
``` python
# Set some constants for training
exp = Experiment(
num_epochs=5,
fp16=False,
device="cuda",
seed=42,
)
exp.compile_experiment(
model_config=config,
callbacks=callbacks,
metrics=metric_list,
main_metrics="accuracy",
)
# Run your experiment with training dataloader and validation dataloader.
exp.fit_loader(train_dl=train_dl, valid_dl=valid_dl)
```
For inference, you can use infer method, which yields output per batch. You can use it as follows
``` python
outputs = []
for op in exp.predict_on_loader(
test_loader=test_dl, path_to_model="./models/model.bin", device="cuda"
):
op = some_post_process_function(op)
outputs.extend(op)
```
If you want to access your experiments history or get as a dataframe. You can do it as follows.
``` python
history = exp.history # This will return a dict
exp.get_logs() # This will return a dataframe constructed from model-history.
```
***
### ***Examples***
* [Image Classification](https://github.com/Atharva-Phatak/torchflare/blob/main/examples/Basic-Tutorials/image_classification.py) using TorchFlare on MNIST dataset.
* [Text Classification](https://github.com/Atharva-Phatak/torchflare/blob/main/examples/Basic-Tutorials/text_classification.py) using Tiny Bert on IMDB dataset
* [Variational Auto-encoders](https://github.com/Atharva-Phatak/torchflare/tree/main/examples/Advanced-Tutorials/autoencoders) to generate MNIST dataset.
* Train [DCGANS](https://github.com/Atharva-Phatak/torchflare/blob/main/examples/Advanced-Tutorials/gans/dcgan.py) to generate MNIST data.
* [Self Supervised learning](https://github.com/Atharva-Phatak/torchflare/blob/main/examples/Advanced-Tutorials/self-supervision/ssl_byol.py) using [Bootstrap your own latent(BYOL)](https://arxiv.org/abs/2006.07733)
***
### ***Contributions***
To contribute please refer to [Contributing Guide](https://github.com/Atharva-Phatak/torchflare/blob/main/.github/CONTRIBUTING.MD)
***
### ***Current Contributors***
***
### ***Author***
* **Atharva Phatak** - [Atharva-Phatak](https://github.com/Atharva-Phatak)
### Citation
Please use this bibtex if you want to cite this repository in your publications:
@misc{TorchFlare,
author = {Atharva Phatak},
title = {TorchFlare: Easy model training and experimentation.},
year = {2020},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/Atharva-Phatak/torchflare}},
}