pytorch多卡训练

参考链接:https://fyubang.com/2019/07/23/distributed-training3/

pytorch多gpu并行训练 - 知乎 (zhihu.com)

PyTorch DistributedDataParallel使用小结_yuanye_yuanye的博客-CSDN博客

GitHub - tczhangzhi/pytorch-distributed: A quickstart and benchmark for pytorch distributed training.

1. 使用DistributedDataParallel并行训练

• 运行命令：CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2 torch_ddp.py
• DistributedDataParallel比DataParallel运行的更快, 然后显存分配的更加均衡。

import torch
import torch.nn as nn
from torch.autograd import Variable
from torch.utils.data import Dataset, DataLoader
import os
from torch.utils.data.distributed import DistributedSampler
# 1) 初始化
torch.distributed.init_process_group(backend="nccl")

input_size = 5
output_size = 2
batch_size = 30
data_size = 90

# 2） 配置每个进程的gpu
local_rank = torch.distributed.get_rank()
torch.cuda.set_device(local_rank)
device = torch.device("cuda", local_rank)

class RandomDataset(Dataset):
    def __init__(self, size, length):
        self.len = length
        self.data = torch.randn(length, size).to('cuda')

    def __getitem__(self, index):
        return self.data[index]

    def __len__(self):
        return self.len

dataset = RandomDataset(input_size, data_size)
# 3）使用DistributedSampler
rand_loader = DataLoader(dataset=dataset,
                         batch_size=batch_size,
                         sampler=DistributedSampler(dataset))

class Model(nn.Module):
    def __init__(self, input_size, output_size):
        super(Model, self).__init__()
        self.fc = nn.Linear(input_size, output_size)

    def forward(self, input):
        output = self.fc(input)
        print("  In Model: input size", input.size(),
              "output size", output.size())
        return output
    
model = Model(input_size, output_size)

# 4) 封装之前要把模型移到对应的gpu
model.to(device)
    
if torch.cuda.device_count() > 1:
    print("Let's use", torch.cuda.device_count(), "GPUs!")
    # 5) 封装
    model = torch.nn.parallel.DistributedDataParallel(model,
                                                      device_ids=[local_rank],
                                                      output_device=local_rank)
   
for data in rand_loader:
    if torch.cuda.is_available():
        input_var = data
    else:
        input_var = data

    output = model(input_var)
    print("Outside: input size", input_var.size(), "output_size", output.size())

• 使用DistributedDataParallel分布式训练，模型中含有BatchNormalization层会出现如下问题：
• 解决方法:1）删除BN层;2)使用SyncBatchNorm层代替