《PyTorch 2.5重磅更新：性能優(yōu)化+新特性》中的一個新特性就是：正式支持在英特爾獨立顯卡上訓(xùn)練模型！

本文將在英特爾 酷睿 Ultra 7 155H自帶的銳炫 集成顯卡上展示使用Pytorch2.5搭建并訓(xùn)練AI模型的全流程。

1搭建開發(fā)環(huán)境

首先，請安裝顯卡驅(qū)動，參考指南：

https://dgpu-docs.intel.com/driver/client/overview.html

并用下面的命令創(chuàng)建并激活名為pytorch_arc的虛擬環(huán)境：

conda create -n pytorch_arc python=3.11  #創(chuàng)建虛擬環(huán)境
conda activate pytorch_arc         #激活虛擬環(huán)境
python -m pip install --upgrade pip    #升級pip到最新版本

接著，安裝Pytorch XPU版；

pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/xpu

最后，執(zhí)行命令，驗證安裝。看到返回結(jié)果為“True”，證明環(huán)境搭建成功！

>>> import torch
>>> torch.xpu.is_available()

2訓(xùn)練ResNet模型

執(zhí)行下載的訓(xùn)練代碼，實現(xiàn)在英特爾 銳炫 集成顯卡上訓(xùn)練ResNet50模型。代碼下載鏈接：

https://gitee.com/Pauntech/Pytorch-2.5

import torchimport torchvision
LR = 0.001DOWNLOAD = TrueDATA = "datasets/cifar10/"
transform = torchvision.transforms.Compose(  [    torchvision.transforms.Resize((224, 224)),    torchvision.transforms.ToTensor(),    torchvision.transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),  ])train_dataset = torchvision.datasets.CIFAR10(  root=DATA,  train=True,  transform=transform,  download=DOWNLOAD,)train_loader = torch.utils.data.DataLoader(dataset=train_dataset, batch_size=128)train_len = len(train_loader)
model = torchvision.models.resnet50()criterion = torch.nn.CrossEntropyLoss()optimizer = torch.optim.SGD(model.parameters(), lr=LR, momentum=0.9)model.train()model = model.to("xpu")criterion = criterion.to("xpu")
print(f"Initiating training")for batch_idx, (data, target) in enumerate(train_loader):  data = data.to("xpu")  target = target.to("xpu")  optimizer.zero_grad()  output = model(data)  loss = criterion(output, target)  loss.backward()  optimizer.step()  if (batch_idx + 1) % 10 == 0:     iteration_loss = loss.item()     print(f"Iteration [{batch_idx+1}/{train_len}], Loss: {iteration_loss:.4f}")torch.save(  {    "model_state_dict": model.state_dict(),    "optimizer_state_dict": optimizer.state_dict(),  },  "checkpoint.pth",)
print("Execution finished")

3總結(jié)

使用PyTorch在英特爾獨立顯卡上訓(xùn)練模型將為AI行業(yè)新增計算硬件選擇！