https://github.com/yusugomori/barbar
Progress bar for deep learning training iterations💈
https://github.com/yusugomori/barbar
deep-learning keras pytorch
Last synced: over 1 year ago
JSON representation
Progress bar for deep learning training iterations💈
- Host: GitHub
- URL: https://github.com/yusugomori/barbar
- Owner: yusugomori
- License: mit
- Created: 2019-04-11T04:06:44.000Z (about 7 years ago)
- Default Branch: master
- Last Pushed: 2019-04-12T02:23:16.000Z (about 7 years ago)
- Last Synced: 2025-03-24T11:56:49.740Z (over 1 year ago)
- Topics: deep-learning, keras, pytorch
- Language: Python
- Homepage:
- Size: 10.7 KB
- Stars: 34
- Watchers: 2
- Forks: 5
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Barbar💈
Progress bar for deep learning training iterations.
## Quick glance
```python
from barbar import Bar
import torch
from torch.utils.data import DataLoader
from torchvision import datasets
mnist_train = datasets.MNIST(root=root,
download=True,
train=True)
train_dataloader = DataLoader(mnist_train,
batch_size=100,
shuffle=True)
model = MLP().to(device)
for epoch in range(epochs):
print('Epoch: {}'.format(epoch+1))
for idx, (x, t) in enumerate(Bar(train_dataloader)):
x, t = x.to(device), t.to(device)
loss, preds = train_step(x, t)
```
```
Epoch: 1
60000/60000: [===============================>] - ETA 0.0s
Epoch: 2
28100/60000: [==============>.................] - ETA 4.1s
```
Barbar works best with PyTorch DataLoader, but it also works with custom DataLoader. Minimal DataLoader example can be written as follows:
```python
class CustomDataLoader(object):
def __init__(self, dataset,
batch_size=100,
shuffle=False,
random_state=None):
self.dataset = list(zip(dataset[0], dataset[1]))
self.batch_size = batch_size
self.shuffle = shuffle
if random_state is None:
random_state = np.random.RandomState(1234)
self.random_state = random_state
self._idx = 0
self._reset()
def __len__(self):
N = len(self.dataset)
b = self.batch_size
return N // b + bool(N % b)
def __iter__(self):
return self
def __next__(self):
if self._idx >= len(self.dataset):
self._reset()
raise StopIteration()
x, y = \
zip(*self.dataset[self._idx:(self._idx + self.batch_size)])
# x = torch.Tensor(x)
# y = torch.LongTensor(y)
self._idx += self.batch_size
return x, y
def _reset(self):
if self.shuffle:
self.dataset = shuffle(self.dataset,
random_state=self.random_state)
self._idx = 0
mnist = datasets.fetch_openml('mnist_784', version=1,)
x, y = mnist.data.astype(np.float32), mnist.target.astype(np.int32)
x = x / 255.
x_train = x[:60000]
y_train = y[:60000]
train_dataloader = CustomDataLoader((x_train, y_train),
batch_size=100,
shuffle=True)
```
## Installation
- **Install Barbar from PyPI (recommended):**
```sh
pip install barbar
```
- **Alternatively: install Barbar from the GitHub source:**
First, clone Barbar using `git`:
```sh
git clone https://github.com/yusugomori/barbar.git
```
Then, `cd` to the Barbar folder and run the install command:
```sh
cd barbar
sudo python setup.py install
```