Files
PSO/test/mnist_torch.py
jung-geun 813433000a feat: modernize PSO and add convergence research
Migrate the package and examples to the tensor-native PyTorch implementation, add benchmark evidence, and add the guarded post-training convergence protocol with TensorBoard progress monitoring and hash-verified recovery.

Constraint: Preserve one-shot official-test sealing and auditable research artifacts

Rejected: Commit local .omc runs and downloaded datasets | multi-gigabyte runtime state is machine-local

Confidence: high

Scope-risk: broad

Not-tested: Production CUDA run on pieroot-server
2026-09-07 22:03:25 +09:00

137 lines
3.8 KiB
Python

"""MNIST dataset gradient baseline (PyTorch standard backprop optimizer, non-PSO)."""
import copy
import numpy as np
import torch
import torch.nn as nn
from torch.utils.data import DataLoader
class MNISTModel(nn.Module):
def __init__(self):
super().__init__()
self.conv1 = nn.Conv2d(1, 64, kernel_size=5)
self.relu1 = nn.ReLU()
self.pool1 = nn.MaxPool2d(2, 2)
self.drop1 = nn.Dropout(0.5)
self.conv2 = nn.Conv2d(64, 128, kernel_size=3)
self.relu2 = nn.ReLU()
self.pool2 = nn.MaxPool2d(2, 2)
self.drop2 = nn.Dropout(0.5)
self.fc1 = nn.Linear(128 * 5 * 5, 2048)
self.relu3 = nn.ReLU()
self.drop3 = nn.Dropout(0.8)
self.fc2 = nn.Linear(2048, 1024)
self.relu4 = nn.ReLU()
self.drop4 = nn.Dropout(0.8)
self.fc3 = nn.Linear(1024, 10)
def forward(self, x):
x = self.drop1(self.pool1(self.relu1(self.conv1(x))))
x = self.pool2(self.relu2(self.conv2(x)))
x = torch.flatten(x, 1)
x = self.drop3(self.relu3(self.fc1(self.drop2(x))))
x = self.drop4(self.relu4(self.fc2(x)))
x = self.fc3(x)
return x
def get_data(download: bool = True):
from torchvision import datasets, transforms
transform = transforms.ToTensor()
train_dataset = datasets.MNIST(
root="./data", train=True, transform=transform, download=download
)
test_dataset = datasets.MNIST(
root="./data", train=False, transform=transform, download=download
)
return train_dataset, test_dataset
def get_device() -> torch.device:
if (
hasattr(torch.backends, "mps")
and torch.backends.mps.is_built()
and torch.backends.mps.is_available()
):
return torch.device("mps")
elif torch.cuda.is_available():
return torch.device("cuda")
return torch.device("cpu")
def main():
torch.manual_seed(42)
np.random.seed(42)
device = get_device()
print(f"Selected device: {device}")
train_dataset, test_dataset = get_data(download=True)
train_loader = DataLoader(train_dataset, batch_size=64, shuffle=True)
test_loader = DataLoader(test_dataset, batch_size=64, shuffle=False)
model = MNISTModel().to(device)
optimizer = torch.optim.Adam(model.parameters(), lr=0.001)
criterion = nn.CrossEntropyLoss()
best_val_loss = float("inf")
best_state = None
for epoch in range(10):
model.train()
for bx, by in train_loader:
bx, by = bx.to(device), by.to(device)
optimizer.zero_grad()
out = model(bx)
loss = criterion(out, by)
loss.backward()
optimizer.step()
model.eval()
val_loss = 0.0
total = 0
with torch.no_grad():
for bx, by in test_loader:
bx, by = bx.to(device), by.to(device)
out = model(bx)
loss = criterion(out, by)
val_loss += loss.item() * bx.size(0)
total += bx.size(0)
val_loss /= total
if val_loss < best_val_loss:
best_val_loss = val_loss
best_state = copy.deepcopy(model.state_dict())
if best_state is not None:
model.load_state_dict(best_state)
model.eval()
test_loss = 0.0
correct = 0
total = 0
with torch.no_grad():
for bx, by in test_loader:
bx, by = bx.to(device), by.to(device)
out = model(bx)
loss = criterion(out, by)
test_loss += loss.item() * bx.size(0)
preds = out.argmax(dim=1)
correct += (preds == by).sum().item()
total += bx.size(0)
test_loss /= total
test_acc = correct / total
print(f"Final test loss: {test_loss:.4f}, accuracy: {test_acc:.4f}")
if __name__ == "__main__":
main()