utils.py

import os
import random

import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F

from gpu_utils import bounce_back_boundary_2d as bounce_back_boundary_2d_cuda


def create_directory(directory: str) -> None:
    if not os.path.exists(directory):
        os.makedirs(directory)


def seed_everything(offset=0):
    torch.manual_seed(1235 + offset)
    torch.backends.cudnn.deterministic = True
    torch.backends.cudnn.benchmark = False
    random.seed(42 + offset)
    np.random.seed(42 + offset)


def seconds_to_human_readable(seconds):
    days, rem = divmod(seconds, 86400)
    hours, rem = divmod(rem, 3600)
    minutes, seconds = divmod(rem, 60)
    if seconds < 1:
        seconds = 1
    locals_ = locals()
    magnitudes_str = (
        "{n} {magnitude}".format(n=int(locals_[magnitude]), magnitude=magnitude)
        for magnitude in ("days", "hours", "minutes", "seconds")
        if locals_[magnitude]
    )
    eta_str = ", ".join(magnitudes_str)
    return eta_str


def train(model, device, train_loader, optimizer, epoch):
    model.train()
    for batch_idx, (data, target) in enumerate(train_loader):
        data, target = data.to(device), target.to(device)
        optimizer.zero_grad()
        output = model(data)
        loss = F.nll_loss(output, target)
        loss.backward()
        optimizer.step()
        if batch_idx % 10 == 0:
            print(
                "Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}".format(
                    epoch,
                    batch_idx * len(data),
                    len(train_loader.dataset),
                    100.0 * batch_idx / len(train_loader),
                    loss.item(),
                )
            )


def bootstrap(model, train_dataset, device, num_batches=10):
    train_loader = torch.utils.data.DataLoader(
        train_dataset, batch_size=64, shuffle=True
    )
    model.train()
    criterion = nn.CrossEntropyLoss()
    optimizer = torch.optim.Adam(model.parameters())
    cnt = 0
    while True:
        cnt += 1
        for batch_idx, (data, target) in enumerate(train_loader):
            data, target = data.to(device), target.to(device)
            optimizer.zero_grad()
            output = model(data)
            loss = criterion(output, target)
            loss.backward()
            optimizer.step()
        if cnt >= num_batches:
            print("Loss after bootstrap: {:.6f}".format(loss.item()))
            return model


def test(model, device, test_loader):
    model.eval()
    test_loss = 0
    correct = 0
    with torch.no_grad():
        for data, target in test_loader:
            data, target = data.to(device), target.to(device)
            output = model(data)
            test_loss += F.nll_loss(output, target, reduction="sum").item()
            pred = output.argmax(dim=1, keepdim=True)
            correct += pred.eq(target.view_as(pred)).sum().item()

    test_loss /= len(test_loader.dataset)

    print(
        "\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n".format(
            test_loss,
            correct,
            len(test_loader.dataset),
            100.0 * correct / len(test_loader.dataset),
        )
    )
    return (test_loss, 100.0 * correct / len(test_loader.dataset))


def train_via_ndes(model, ndes, device, test_dataset, model_name):
    model.eval()
    test_loader = torch.utils.data.DataLoader(
        test_dataset, batch_size=1000, shuffle=True
    )

    model = ndes.run(lambda x: test(x, device, test_loader))
    test(model, device, test_loader)
    torch.save({"state_dict": model.state_dict()}, f"{model_name}_{ndes.start}.pth.tar")


def train_via_gradient(
    model, criterion, optimizer, train_dataset, test_dataset, num_epoch, device
):
    model.train()
    train_loader = torch.utils.data.DataLoader(
        train_dataset, batch_size=64, shuffle=True
    )
    test_loader = torch.utils.data.DataLoader(
        test_dataset, batch_size=1000, shuffle=True
    )

    for epoch in range(1, num_epoch + 1):
        train(model, device, train_loader, optimizer, epoch)
        test(model, device, test_loader)


# @profile
def bounce_back_boundary_1d(x, lower, upper):
    """TODO

    Examples:
    >>> a = torch.tensor([-2429.4529, 10, 580.3583, -10, 1316.1814, 0, 0])
    >>> lower = torch.ones(7) * -2000
    >>> upper = torch.ones(7) * 500
    >>> bounce_back_boundary_1d(a, lower, upper)
    tensor([-1570.5471,    10.0000,   419.6417,   -10.0000,  -316.1814,
    0.0000, 0.0000])
    """
    is_lower_boundary_ok = x >= lower
    is_upper_boundary_ok = x <= upper
    if is_lower_boundary_ok.all() and is_upper_boundary_ok.all():
        return x
    delta = upper - lower
    x = torch.where(is_lower_boundary_ok, x, lower + ((lower - x) % delta))
    x = torch.where(is_upper_boundary_ok, x, upper - (((upper - x) * -1) % delta))
    return x


# @profile
def bounce_back_boundary_2d(x, lower, upper):
    lower = lower[0]
    upper = upper[0]
    delta = upper - lower
    return bounce_back_boundary_2d_cuda(x, lower, upper, delta)