# https://github.com/pytorch/examples/blob/main/distributed/rpc/pipeline/main.py
import os
import threading
import time
import torch
import torch.nn as nn
import torch.distributed.autograd as dist_autograd
import torch.distributed.rpc as rpc
import torch.optim as optim
from torch.distributed.optim import DistributedOptimizer
from torch.distributed.rpc import RRef
from torchvision.models.resnet import Bottleneck
os.environ['MASTER_ADDR'] = 'XXX.XXX.XXX.XXX' # 指定master ip地址
os.environ['MASTER_PORT'] = '7856' # 指定master 端口號(hào)
#########################################################
#           Define Model Parallel ResNet50              #
#########################################################
# In order to split the ResNet50 and place it on two different workers, we
# implement it in two model shards. The ResNetBase class defines common
# attributes and methods shared by two shards. ResNetShard1 and ResNetShard2
# contain two partitions of the model layers respectively.
num_classes = 1000
def conv1x1(in_planes, out_planes, stride=1):
    """1x1 convolution"""
    return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)
class ResNetBase(nn.Module):
    def __init__(self, block, inplanes, num_classes=1000,
                 groups=1, width_per_group=64, norm_layer=None):
        super(ResNetBase, self).__init__()
        self._lock = threading.Lock()
        self._block = block
        self._norm_layer = nn.BatchNorm2d
        self.inplanes = inplanes
        self.dilation = 1
        self.groups = groups
        self.base_width = width_per_group
    def _make_layer(self, planes, blocks, stride=1):
        norm_layer = self._norm_layer
        downsample = None
        previous_dilation = self.dilation
        if stride != 1 or self.inplanes != planes * self._block.expansion:
            downsample = nn.Sequential(
                conv1x1(self.inplanes, planes * self._block.expansion, stride),
                norm_layer(planes * self._block.expansion),
            )
        layers = []
        layers.append(self._block(self.inplanes, planes, stride, downsample, self.groups,
                                  self.base_width, previous_dilation, norm_layer))
        self.inplanes = planes * self._block.expansion
        for _ in range(1, blocks):
            layers.append(self._block(self.inplanes, planes, groups=self.groups,
                                      base_width=self.base_width, dilation=self.dilation,
                                      norm_layer=norm_layer))
        return nn.Sequential(*layers)
    def parameter_rrefs(self):
        r"""
        Create one RRef for each parameter in the given local module, and return a
        list of RRefs.
        """
        return [RRef(p) for p in self.parameters()]
class ResNetShard1(ResNetBase):
    """
    The first part of ResNet.
    """
    def __init__(self, device, *args, **kwargs):
        super(ResNetShard1, self).__init__(
            Bottleneck, 64, num_classes=num_classes, *args, **kwargs)
        self.device = device
        self.seq = nn.Sequential(
            nn.Conv2d(3, self.inplanes, kernel_size=7, stride=2, padding=3, bias=False),
            self._norm_layer(self.inplanes),
            nn.ReLU(inplace=True),
            nn.MaxPool2d(kernel_size=3, stride=2, padding=1),
            self._make_layer(64, 3),
            self._make_layer(128, 4, stride=2)
        ).to(self.device)
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
            elif isinstance(m, nn.BatchNorm2d):
                nn.init.ones_(m.weight)
                nn.init.zeros_(m.bias)
    def forward(self, x_rref):
        x = x_rref.to_here().to(self.device)
        with self._lock:
            out =  self.seq(x)
        return out.cpu()
class ResNetShard2(ResNetBase):
    """
    The second part of ResNet.
    """
    def __init__(self, device, *args, **kwargs):
        super(ResNetShard2, self).__init__(
            Bottleneck, 512, num_classes=num_classes, *args, **kwargs)
        self.device = device
        self.seq = nn.Sequential(
            self._make_layer(256, 6, stride=2),
            self._make_layer(512, 3, stride=2),
            nn.AdaptiveAvgPool2d((1, 1)),
        ).to(self.device)
        self.fc =  nn.Linear(512 * self._block.expansion, num_classes).to(self.device)
    def forward(self, x_rref):
        x = x_rref.to_here().to(self.device)
        with self._lock:
            out = self.fc(torch.flatten(self.seq(x), 1))
        return out.cpu()
class DistResNet50(nn.Module):
    """
    Assemble two parts as an nn.Module and define pipelining logic
    """
    def __init__(self, split_size, workers, *args, **kwargs):
        super(DistResNet50, self).__init__()
        self.split_size = split_size
        # Put the first part of the ResNet50 on workers[0]
        self.p1_rref = rpc.remote(
            workers[0],
            ResNetShard1,
            args = ("cuda:0",) + args,
            kwargs = kwargs
        )
        # Put the second part of the ResNet50 on workers[1]
        self.p2_rref = rpc.remote(
            workers[1],
            ResNetShard2,
            args = ("cpu",) + args,
            kwargs = kwargs
        )
    def forward(self, xs):
        # Split the input batch xs into micro-batches, and collect async RPC
        # futures into a list
        out_futures = []
        for x in iter(xs.split(self.split_size, dim=0)):
            x_rref = RRef(x)
            y_rref = self.p1_rref.remote().forward(x_rref)
            print(y_rref)
            z_fut = self.p2_rref.rpc_async().forward(y_rref)
            print(z_fut)
            out_futures.append(z_fut)
        # collect and cat all output tensors into one tensor.
        return torch.cat(torch.futures.wait_all(out_futures))
    def parameter_rrefs(self):
        remote_params = []
        remote_params.extend(self.p1_rref.remote().parameter_rrefs().to_here())
        remote_params.extend(self.p2_rref.remote().parameter_rrefs().to_here())
        return remote_params
#########################################################
#                   Run RPC Processes                   #
#########################################################
num_batches = 3
batch_size = 8
image_w = 128
image_h = 128
if __name__=="__main__":
    options = rpc.TensorPipeRpcBackendOptions(num_worker_threads=256, rpc_timeout=300)
    # 初始化主節(jié)點(diǎn)的RPC連接
    rpc.init_rpc("master", rank=0, world_size=2, rpc_backend_options=options)
    for num_split in [1,2]:
        tik = time.time()
        model = DistResNet50(num_split, ["master", "worker"])
        loss_fn = nn.MSELoss()
        opt = DistributedOptimizer(
            optim.SGD,
            model.parameter_rrefs(),
            lr=0.05,
        )
        one_hot_indices = torch.LongTensor(batch_size) \
            .random_(0, num_classes) \
            .view(batch_size, 1)
        for i in range(num_batches):
            print(f"Processing batch {i}")
            # generate random inputs and labels
            inputs = torch.randn(batch_size, 3, image_w, image_h)
            labels = torch.zeros(batch_size, num_classes) \
                .scatter_(1, one_hot_indices, 1)
            with dist_autograd.context() as context_id:
                outputs = model(inputs)
                dist_autograd.backward(context_id, [loss_fn(outputs, labels)])
                opt.step(context_id)
        tok = time.time()
        print(f"number of splits = {num_split}, execution time = {tok - tik}")
    # 關(guān)閉RPC連接
    rpc.shutdown()

# https://github.com/pytorch/examples/blob/main/distributed/rpc/pipeline/main.py
import os
import threading
import time
from functools import wraps
import torch
import torch.nn as nn
import torch.distributed.rpc as rpc
from torch.distributed.rpc import RRef
from torchvision.models.resnet import Bottleneck
os.environ['MASTER_ADDR'] = 'XXX.XXX.XXX.XXX' # 指定master 端口號(hào)
os.environ['MASTER_PORT'] = '7856' # 指定master 端口號(hào)
#########################################################
#           Define Model Parallel ResNet50              #
#########################################################
# In order to split the ResNet50 and place it on two different workers, we
# implement it in two model shards. The ResNetBase class defines common
# attributes and methods shared by two shards. ResNetShard1 and ResNetShard2
# contain two partitions of the model layers respectively.
num_classes = 1000
def conv1x1(in_planes, out_planes, stride=1):
    """1x1 convolution"""
    return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)
class ResNetBase(nn.Module):
    def __init__(self, block, inplanes, num_classes=1000,
                 groups=1, width_per_group=64, norm_layer=None):
        super(ResNetBase, self).__init__()
        self._lock = threading.Lock()
        self._block = block
        self._norm_layer = nn.BatchNorm2d
        self.inplanes = inplanes
        self.dilation = 1
        self.groups = groups
        self.base_width = width_per_group
    def _make_layer(self, planes, blocks, stride=1):
        norm_layer = self._norm_layer
        downsample = None
        previous_dilation = self.dilation
        if stride != 1 or self.inplanes != planes * self._block.expansion:
            downsample = nn.Sequential(
                conv1x1(self.inplanes, planes * self._block.expansion, stride),
                norm_layer(planes * self._block.expansion),
            )
        layers = []
        layers.append(self._block(self.inplanes, planes, stride, downsample, self.groups,
                                  self.base_width, previous_dilation, norm_layer))
        self.inplanes = planes * self._block.expansion
        for _ in range(1, blocks):
            layers.append(self._block(self.inplanes, planes, groups=self.groups,
                                      base_width=self.base_width, dilation=self.dilation,
                                      norm_layer=norm_layer))
        return nn.Sequential(*layers)
    def parameter_rrefs(self):
        r"""
        Create one RRef for each parameter in the given local module, and return a
        list of RRefs.
        """
        return [RRef(p) for p in self.parameters()]
class ResNetShard2(ResNetBase):
    """
    The second part of ResNet.
    """
    def __init__(self, device, *args, **kwargs):
        super(ResNetShard2, self).__init__(
            Bottleneck, 512, num_classes=num_classes, *args, **kwargs)
        self.device = device
        self.seq = nn.Sequential(
            self._make_layer(256, 6, stride=2),
            self._make_layer(512, 3, stride=2),
            nn.AdaptiveAvgPool2d((1, 1)),
        ).to(self.device)
        self.fc =  nn.Linear(512 * self._block.expansion, num_classes).to(self.device)
    def forward(self, x_rref):
        x = x_rref.to_here().to(self.device)
        print(x)
        with self._lock:
            out = self.fc(torch.flatten(self.seq(x), 1))
        return out.cpu()
#########################################################
#                   Run RPC Processes                   #
#########################################################
if __name__=="__main__":
    options = rpc.TensorPipeRpcBackendOptions(num_worker_threads=256, rpc_timeout=300)
    # 初始化工作節(jié)點(diǎn)的RPC連接
    rpc.init_rpc("worker", rank=1, world_size=2, rpc_backend_options=options)
    # 等待主節(jié)點(diǎn)的調(diào)用
    rpc.shutdown()