prototorch_models/examples/lvqmln_iris.py

107 lines
2.6 KiB
Python
Raw Permalink Normal View History

2021-05-17 15:03:37 +00:00
"""LVQMLN example using all four dimensions of the Iris dataset."""
2021-05-30 22:52:16 +00:00
2021-05-21 15:55:55 +00:00
import argparse
import warnings
2021-05-17 15:03:37 +00:00
import prototorch as pt
2021-05-17 15:03:37 +00:00
import pytorch_lightning as pl
import torch
from lightning_fabric.utilities.seed import seed_everything
from prototorch.models import (
LVQMLN,
PruneLoserPrototypes,
VisSiameseGLVQ2D,
)
from pytorch_lightning.utilities.warnings import PossibleUserWarning
from torch.utils.data import DataLoader
warnings.filterwarnings("ignore", category=PossibleUserWarning)
warnings.filterwarnings("ignore", category=UserWarning)
2021-05-17 15:03:37 +00:00
2021-05-21 15:55:55 +00:00
class Backbone(torch.nn.Module):
2021-05-21 15:55:55 +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.activation = torch.nn.Sigmoid()
def forward(self, x):
x = self.activation(self.dense1(x))
out = self.activation(self.dense2(x))
return out
2021-05-17 15:03:37 +00:00
if __name__ == "__main__":
2021-05-21 15:55:55 +00:00
# Command-line arguments
parser = argparse.ArgumentParser()
parser.add_argument("--gpus", type=int, default=0)
parser.add_argument("--fast_dev_run", type=bool, default=False)
2021-05-21 15:55:55 +00:00
args = parser.parse_args()
2021-05-17 15:03:37 +00:00
# Dataset
train_ds = pt.datasets.Iris()
# Reproducibility
seed_everything(seed=42)
2021-05-17 15:03:37 +00:00
# Dataloaders
train_loader = DataLoader(train_ds, batch_size=150)
2021-05-17 15:03:37 +00:00
# Hyperparameters
hparams = dict(
distribution=[3, 4, 5],
2021-05-17 15:03:37 +00:00
proto_lr=0.001,
bb_lr=0.001,
)
# Initialize the backbone
backbone = Backbone()
# Initialize the model
model = LVQMLN(
2021-05-17 15:03:37 +00:00
hparams,
prototypes_initializer=pt.initializers.SSCI(
train_ds,
transform=backbone,
),
2021-05-17 15:03:37 +00:00
backbone=backbone,
)
# Callbacks
vis = VisSiameseGLVQ2D(
2021-05-17 15:03:37 +00:00
data=train_ds,
map_protos=False,
border=0.1,
resolution=500,
axis_off=True,
)
pruning = PruneLoserPrototypes(
threshold=0.01,
idle_epochs=20,
prune_quota_per_epoch=2,
frequency=10,
verbose=True,
)
2021-05-17 15:03:37 +00:00
# Setup trainer
trainer = pl.Trainer(
accelerator="cuda" if args.gpus else "cpu",
devices=args.gpus if args.gpus else "auto",
fast_dev_run=args.fast_dev_run,
callbacks=[
vis,
pruning,
],
log_every_n_steps=1,
max_epochs=1000,
detect_anomaly=True,
2021-05-21 15:55:55 +00:00
)
2021-05-17 15:03:37 +00:00
# Training loop
trainer.fit(model, train_loader)