prototorch_models/examples/siamese_glvq_iris.py

62 lines
1.6 KiB
Python
Raw Normal View History

2021-04-27 12:35:17 +00:00
"""Siamese GLVQ example using all four dimensions of the Iris dataset."""
2021-05-07 13:25:04 +00:00
import prototorch as pt
2021-04-27 12:35:17 +00:00
import pytorch_lightning as pl
import torch
2021-04-27 12:35:17 +00:00
class Backbone(torch.nn.Module):
"""Two fully connected layers with ReLU activation."""
2021-04-27 12:35:17 +00:00
def __init__(self, input_size=4, hidden_size=10, latent_size=2):
super().__init__()
self.input_size = input_size
self.hidden_size = hidden_size
self.latent_size = latent_size
self.dense1 = torch.nn.Linear(self.input_size, self.hidden_size)
self.dense2 = torch.nn.Linear(self.hidden_size, self.latent_size)
self.relu = torch.nn.ReLU()
def forward(self, x):
x = self.relu(self.dense1(x))
out = self.relu(self.dense2(x))
return out
2021-04-27 12:35:17 +00:00
if __name__ == "__main__":
# Dataset
2021-05-11 15:22:02 +00:00
train_ds = pt.datasets.Iris()
2021-04-27 12:35:17 +00:00
# Reproducibility
pl.utilities.seed.seed_everything(seed=2)
2021-04-27 12:35:17 +00:00
# Dataloaders
2021-05-07 13:25:04 +00:00
train_loader = torch.utils.data.DataLoader(train_ds,
num_workers=0,
batch_size=150)
2021-04-27 12:35:17 +00:00
# Hyperparameters
hparams = dict(
2021-05-11 14:15:08 +00:00
distribution=[1, 2, 3],
proto_lr=0.01,
bb_lr=0.01,
2021-04-27 12:35:17 +00:00
)
# Initialize the model
2021-05-07 13:25:04 +00:00
model = pt.models.SiameseGLVQ(
hparams,
prototype_initializer=pt.components.SMI(train_ds),
backbone_module=Backbone,
)
2021-04-27 12:35:17 +00:00
# Model summary
print(model)
# Callbacks
2021-05-11 15:22:02 +00:00
vis = pt.models.VisSiameseGLVQ2D(data=train_ds, border=0.1)
2021-04-27 12:35:17 +00:00
# Setup trainer
2021-05-13 13:22:01 +00:00
trainer = pl.Trainer(max_epochs=100, callbacks=[vis], gpus=-1)
2021-04-27 12:35:17 +00:00
# Training loop
trainer.fit(model, train_loader)