ROCm 5.4.3 在 ubuntu 20.04 上安装并运行示例

0. 官方说明：

AMD Documentation - Portal

AMD Documentation - Portal

按照官方文档，操作到底：

prerequisite 中也有操作；

显卡类型：镭7  Radeon vii，1200元闲鱼

1. 前置条件

2. 安装 

2.1

2.2

2.3

部分安装命令，前面的页面也有分组之类的，可能不在这里，后半部分：

  189  uname -srmv
  190  lspci | grep -i Display
  191  sudo lspci | grep -i Display
  192  lspci | grep -i display
  193  lspci
  194  sudo apt list --installed | egrep 'wget\|gnupg2|gawk|curl'
  195  sudo apt-get update && sudo apt-get install wget gnupg2 gawk curl
  196  ls /opt/rocm -all
  197  ls /opt/rocm-5.4.3/ -all
  198  ls /opt/rocm-5.4.3 -all
  199  ls /opt/rocm -all
  200  ls /etc/alternatives/rocm -all
  201  cd /opt/rocm-5.4.3/
  202  ls
  203  ls opencl/
  204  ls opencl/lib/
  205  ls
  206  ls lib
  207  ls lib/libOpenCL.so -all
  208  ls lib/libOpenCL.so.a -all
  209  ls lib/libOpenCL.so.1 -all
  210  cd
  211  sudo amdgpu-install --usecase=hiplibsdk,rocm
  212  sudo amdgpu-install --list-usecase
  213  sudo dpkg -l | grep linux-headers
  214  sudo dpkg -l | grep linux-modules-extra
  215  sudo apt install linux-headers-`uname -r` linux-modules-extra-`uname -r`
  216  ls
  217  mkdir installrocm
  218  cd installrocm/
  219  curl -fsSL https://repo.radeon.com/rocm/rocm.gpg.key | sudo gpg --dearmor -o /etc/apt/trusted.gpg.d/rocm-keyring.gpg
  220  echo 'deb [arch=amd64 signed-by=/etc/apt/trusted.gpg.d/rocm-keyring.gpg] https://repo.radeon.com/amdgpu/5.4.3/ubuntu focal main' | sudo tee /etc/apt/sources.list.d/amdgpu.list
  221  sudo apt-get update
  222  sudo apt install amdgpu-dkms
  223  sudo reboot
  224  ls
  225  cd ex
  226  mkdir glfw3_ex
  227  cd glfw3_ex/
  228  ls
  229  wget https://github.com/glfw/glfw/releases/download/3.3.8/glfw-3.3.8.zip
  230  ls
  231  unzip
  232  unzip glfw-3.3.8.zip
  233  ls
  234  cd glfw-3.3.8/
  235  ls
  236  mkdir build
  237  cd build/
  238  cmake ..
  239  sudo apt-get install libxrandr-dev
  240  cmake ..
  241  sudo apt-get install libsdl2-dev
  242  cmake ..
  243  ls
  244  make -j
  
  

3.  运行示例

3.1 下载示例：

git clone https://github.com/amd/rocm-examples.git

3.2 

mkdir build
cd build
cmake ..
make -j

如果缺少glfw3之类的，可以参考：

Ubuntu18.04安装glfw3.3_qq_38196982的博客-CSDN博客

3.3 run:

运行rocBLAS 的 gemm_stride...:

运行示例的同时，执行rocm-smi:

发现问题升高，gpu%占用上升 

4. 安装pytorch

pytorch 官方目前支持到 rocm 5.4.2, 虽然没有读过官方文档，只好假设 rocm是向前兼容的，于是直接安装 这个版本的pytorch即可：

rocman@Radeonnvii-host:~$ sudo pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm5.4.2

4.2 运行示例：

gpu版本的 mnist 代码：

from __future__ import print_function
import argparse
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
from torch.optim.lr_scheduler import StepLR


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. * 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. * 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=14, metavar='N',
                        help='number of epochs to train (default: 14)')
    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()
    use_mps = not args.no_mps and torch.backends.mps.is_available()

    torch.manual_seed(args.seed)

    if use_cuda:
        device = torch.device("cuda")
    elif use_mps:
        device = torch.device("mps")
    else:
        device = torch.device("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('../data', train=True, download=True,
                       transform=transform)
    dataset2 = datasets.MNIST('../data', 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)
    for epoch in range(1, args.epochs + 1):
        train(args, model, device, train_loader, optimizer, epoch)
        test(model, device, test_loader)
        scheduler.step()

    if args.save_model:
        torch.save(model.state_dict(), "mnist_cnn.pt")


if __name__ == '__main__':
    main()

rocm 5.4.3 安装 pytorch-rocm 5.4.2 发现跑出来的结果不正确，跟没有训练是一样的：

 

用cuda能跑正确：

 

虽然可以查找到设备：