3.2 A Pytorch example

from __future__ import print_function

import argparse

import os

import torch

import torch.nn as nn

import torch.nn.functional as F

import torch.optim as optim

from torch.optim.lr_scheduler import StepLR

from torchvision import datasets, transforms

from flavor.cook.utils import SaveInfoJson, SetEvent, WaitEvent # import Flavor

class Net(nn.Module):

    def __init__(self):

        super(Net, self).__init__()

        self.conv1 = nn.Conv2d(1, 32, 3, 1)

        self.conv2 = nn.Conv2d(32, 64, 3, 1)

        self.dropout1 = nn.Dropout(0.25)

        self.dropout2 = nn.Dropout(0.5)

        self.fc1 = nn.Linear(9216, 128)

        self.fc2 = nn.Linear(128, 10)

    def forward(self, x):

        x = self.conv1(x)

        x = F.relu(x)

        x = self.conv2(x)

        x = F.relu(x)

        x = F.max_pool2d(x, 2)

        x = self.dropout1(x)

        x = torch.flatten(x, 1)

        x = self.fc1(x)

        x = F.relu(x)

        x = self.dropout2(x)

        x = self.fc2(x)

        output = F.log_softmax(x, dim=1)

        return output

def train(args, 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 % args.log_interval == 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(),

                )

            )

            if args.dry_run:

                break

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()  # sum up batch loss

            pred = output.argmax(dim=1, keepdim=True)  # get the index of the max log-probability

            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 100.0 * correct / len(test_loader.dataset)

def main():

    # Training settings

    parser = argparse.ArgumentParser(description="PyTorch MNIST Example")

    parser.add_argument(

        "--batch-size",

        type=int,

        default=64,

        metavar="N",

        help="input batch size for training (default: 64)",

    )

    parser.add_argument(

        "--test-batch-size",

        type=int,

        default=1000,

        metavar="N",

        help="input batch size for testing (default: 1000)",

    )

    parser.add_argument(

        "--epochs",

        type=int,

        default=300,

        metavar="N",

        help="number of epochs to train (default: 300)",

    )

    parser.add_argument(

        "--lr", type=float, default=1.0, metavar="LR", help="learning rate (default: 1.0)"

    )

    parser.add_argument(

        "--gamma",

        type=float,

        default=0.7,

        metavar="M",

        help="Learning rate step gamma (default: 0.7)",

    )

    parser.add_argument(

        "--no-cuda", action="store_true", default=False, help="disables CUDA training"

    )

    parser.add_argument(

        "--no-mps", action="store_true", default=False, help="disables macOS GPU training"

    )

    parser.add_argument(

        "--dry-run", action="store_true", default=False, help="quickly check a single pass"

    )

    parser.add_argument("--seed", type=int, default=1, metavar="S", help="random seed (default: 1)")

    parser.add_argument(

        "--log-interval",

        type=int,

        default=10,

        metavar="N",

        help="how many batches to wait before logging training status",

    )

    parser.add_argument(

        "--save-model", action="store_true", default=False, help="For Saving the current Model"

    )

    args = parser.parse_args()

    use_cuda = not args.no_cuda and torch.cuda.is_available()

    torch.manual_seed(args.seed)

    device = torch.device("cuda" if use_cuda else "cpu")

    train_kwargs = {"batch_size": args.batch_size}

    test_kwargs = {"batch_size": args.test_batch_size}

    if use_cuda:

        cuda_kwargs = {"num_workers": 1, "pin_memory": True, "shuffle": True}

        train_kwargs.update(cuda_kwargs)

        test_kwargs.update(cuda_kwargs)

    transform = transforms.Compose(

        [transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,))]

    )

    dataset1 = datasets.MNIST(

        os.environ["INPUT_PATH"], train=True, download=True, transform=transform

    ) # Use an environment variable os.environ["INPUT_PATH"] instead of "./MNIST"

    dataset2 = datasets.MNIST(os.environ["INPUT_PATH"], train=False, transform=transform)

    train_loader = torch.utils.data.DataLoader(dataset1, **train_kwargs)

    test_loader = torch.utils.data.DataLoader(dataset2, **test_kwargs)

    model = Net().to(device)

    optimizer = optim.Adadelta(model.parameters(), lr=args.lr)

    scheduler = StepLR(optimizer, step_size=1, gamma=args.gamma)

    # Tell the server that all preparations for training have been completed.

    SetEvent("TrainInitDone")

    epoch=-1

    while True:

        epoch+=1

        # Wait for the server

        WaitEvent("TrainStarted")

        # Load checkpoint sent from the server

        if epoch != 0 or os.path.exists(os.environ["GLOBAL_MODEL_PATH"]): # /weights/weights.pth

            model.load_state_dict(torch.load(os.environ["GLOBAL_MODEL_PATH"])["state_dict"])

        # Verify the performance of the global model before training

        precision = test(model, device, test_loader)

        # Save information that the server needs to know

        output_dict = {}

        output_dict["metadata"] = {"epoch": epoch, "datasetSize": len(dataset1), "importance": 1.0}

        output_dict["metrics"] = {

            "precision": precision,

            "basic/confusion_tp": -1,  # If N/A or you don't want to track, fill in -1.

            "basic/confusion_fp": -1,

            "basic/confusion_fn": -1,

            "basic/confusion_tn": -1,

        }

        SaveInfoJson(output_dict) # Save json to the server

        train(args, model, device, train_loader, optimizer, epoch)

        scheduler.step()

        # Save checkpoint

        torch.save({"state_dict": model.state_dict()}, os.environ["LOCAL_MODEL_PATH"])

        # Tell the server that this round of training work has ended.

        SetEvent("TrainFinished")

if __name__ == "__main__":

    main()

FROM pytorch/pytorch:1.12.0-cuda11.3-cudnn8-runtime

COPY . /app

workdir /app

RUN pip install -r requirements.txt

ENV PROCESS="python main.py"

CMD flavor-fl -m "${PROCESS}"

torch

torchvision

https://github.com/ailabstw/FLaVor/archive/refs/heads/release/stable.zip