Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/jamormoussa/torch_utils
torch_utils is a PyTorch extension designed for training and building deep learning models.
https://github.com/jamormoussa/torch_utils
python pytorch
Last synced: 5 days ago
JSON representation
torch_utils is a PyTorch extension designed for training and building deep learning models.
- Host: GitHub
- URL: https://github.com/jamormoussa/torch_utils
- Owner: JamorMoussa
- Created: 2024-05-14T22:30:41.000Z (7 months ago)
- Default Branch: main
- Last Pushed: 2024-05-27T23:19:39.000Z (7 months ago)
- Last Synced: 2024-12-16T04:17:28.195Z (9 days ago)
- Topics: python, pytorch
- Language: Python
- Homepage:
- Size: 33.2 KB
- Stars: 3
- Watchers: 2
- Forks: 0
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# torch_utils
**torch_utils** is a PyTorch extension designed for training and building deep learning models.
## Example
```python
import torch, torch.nn as nn
import torch.optim as optim
from torch.utils.data import TensorDataset, random_split
# import the torch_utils
import torch_utils as ts
```
For this example, we're going to use a simple data. which is generated randomly.
```python
X = torch.rand(100, 3)
y = torch.mm(X, torch.Tensor([1, 2, 3]).unsqueeze(0).t())
dataset = TensorDataset(X, y)
```
Let build a simple model, just a single `Linear` layer.
```python
model = nn.Sequential(
nn.Linear(3, 1)
)
```
Now, let create an instance of our `Trainer` class. The `Trainer` will charge of training of the model.
```python
trainer = ts.train.Trainer()
```
```
Trainer(
(configs): TrainerConfigs(max_iters=10, train_batch=64, test_batch=64, shuffle_data=True, device=device(type='cpu'))
)
```
Let's change some configurations. For example, change the `max_ietrs` to be 100.
```python
trainer.configs.max_iters = 100
```
```python
train_set, test_set = random_split(dataset, [0.8, 0.2])
```
The trainer support `DataContainer`, takes the train, and the test sets.
```python
data_container = ts.data.DataContainer(train_set, test_set)
```
aloso the trainer takes the `OptimizerContainer` rather than the torch `Optimizer`:
```python
opt = ts.optim.OptimizerContainer(optim.Adam(model.parameters(), lr=0.01))
```
Now, let's compile the trainer. Provide all building blocks needed to train the model, such as optimizer, loss functio, datasets.
```python
trainer.compile(
model= model,
opt_container= opt,
loss_fn= nn.MSELoss(),
data_container= data_container
)
```
Finally, train the mode, using the `trainer.train()`
```python
trainer.train()
```
```
100%|████████████████████████████████████████| 100/100 [00:00<00:00, 325.21it/s]
```