pytorch进行单机多卡训练
首先得要了解本机有几张卡可以用：nvidia-smi模型用DataParallel包装一下：device_ids = [1, 2, 3] # 可用GPUmodel_t = torchvision.models.resnet152()fc = model_t.fc.in_featuresmodel_t.fc = nn.Linear(fc, len(df_dict))model_t = torch.nn.DataParallel(model_t, device_ids=device_id.
