dist_context.py

#   Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License

import copy
from collections import defaultdict
import paddle.fluid
from paddle.fluid import framework
from paddle.fluid.framework import get_flags, set_flags
from paddle.fluid import core
from paddle.distributed.passes import PassContext
from .dist_attribute import TensorDistributedAttribute
from .dist_attribute import OperatorDistributedAttribute
from .dist_tensor import DistributedTensor
from .dist_op import DistributedOperator
from .process_mesh import ProcessMesh

# There always exists a default context for user. And user can set it to another one.
_g_default_distributed_context = None


def get_default_distributed_context():
    global _g_default_distributed_context
    if _g_default_distributed_context is None:
        dist_context = DistributedContext()
        set_default_distributed_context(dist_context)
    return _g_default_distributed_context


def set_default_distributed_context(dist_context):
    global _g_default_distributed_context
    _g_default_distributed_context = dist_context


def _node_id(node):
    return (node.node.graph_id(), node.node.id())


class DistributedContext:
    """
    DistributedContext is used to collect related distributed information for program and graph.
    One auto-parallel run should use its own DistributedContext to avoid interfering other run.
    """

    def __init__(self,
                 serial_main_prog=None,
                 serial_startup_prog=None,
                 serial_optimizer=None,
                 serial_loss=None,
                 feed_vars=None,
                 fetch_vars=None,
                 strategy=None):
        # Data members related to original programs (unchanged)
        self._original_serial_main_program = serial_main_prog
        self._original_serial_startup_program = serial_startup_prog
        self._original_serial_loss = serial_loss
        self._original_serial_optimizer = serial_optimizer
        if self._original_serial_main_program is None:
            self._original_serial_main_program = paddle.fluid.default_main_program(
            )
        if self._original_serial_startup_program is None:
            self._original_serial_startup_program = paddle.fluid.default_startup_program(
            )

        # Data members related to programs (changed)
        self._serial_main_program = None
        self._serial_startup_program = None
        self._serial_loss = serial_loss
        self._serial_optimizer = serial_optimizer
        self._serial_feed_vars = feed_vars
        self._serial_fetch_vars = fetch_vars

        # Data members related to the program
        self._dist_tensors_for_program = {}
        self._dist_ops_for_program = {}
        self._block_state = BlockState()

        # Data members related to the graph
        self._serial_graph = None
        self._dist_tensors_for_graph = {}
        self._dist_ops_for_graph = {}
        self._node_id_to_tensor_id = {}
        self._node_id_to_op_id = {}

        # Data members related to the distributed programs
        # Distributed programs
        self._dist_main_programs = {}
        self._dist_startup_programs = {}

        # Distributed Strategy
        self._strategy = strategy

        # Pass Context
        self._pass_context = PassContext()

        # Distributed Operator Context
        self._dist_op_context = DistributedOperatorContext()

        # Other data members
        self._process_meshes = []
        self._serial_ordered_tensor_nodes = []
        self._serial_ordered_op_nodes = []
        self._serial_ordered_nodes = []
        # self._tensor_id_to_tensor_node_ids = {}

        self._is_initialized = False

        # flag whether scale gradient with dp size
        self._gradient_scale = True

    @property
    def serial_main_program(self):
        return self._serial_main_program

    @serial_main_program.setter
    def serial_main_program(self, program):
        # if self._serial_main_program:
        #     print("WARNING: The program attached to this distributed context will be replaced by the new one.")
        self._original_serial_main_program = program
        self._serial_main_program = program

    @property
    def serial_startup_program(self):
        return self._serial_startup_program

    @property
    def serial_loss(self):
        return self._serial_loss

    @property
    def serial_optimizer(self):
        return self._serial_optimizer

    @property
    def serial_feed_vars(self):
        return self._serial_feed_vars

    @property
    def serial_fetch_vars(self):
        return self._serial_fetch_vars

    @property
    def strategy(self):
        return self._strategy

    @property
    def serial_graph(self):
        return self._serial_graph

    @property
    def serial_ordered_nodes(self):
        return self._serial_ordered_nodes

    @property
    def process_meshes(self):
        return self._process_meshes

    @property
    def pass_context(self):
        return self._pass_context

    @property
    def dist_op_context(self):
        return self._dist_op_context

    @property
    def block_state(self):
        return self._block_state

    @property
    def dist_main_programs(self):
        return self._dist_main_programs

    @property
    def dist_startup_programs(self):
        return self._dist_startup_programs

    @property
    def has_annotation(self):
        return len(self._dist_tensors_for_program) or len(
            self._dist_ops_for_program)

    @property
    def gradient_scale(self):
        return self._gradient_scale

    @gradient_scale.setter
    def gradient_scale(self, gs):
        self._gradient_scale = gs

    def initialize(self):
        if not self._is_initialized:
            self._serial_main_program = self._original_serial_main_program.clone(
            )
            self._serial_startup_program = self._original_serial_startup_program.clone(
            )
            # self._serial_main_program = self._original_serial_main_program
            # self._serial_startup_program = self._original_serial_startup_program
            if self._original_serial_loss:
                self._serial_loss = self._serial_main_program.global_block(
                ).vars[self._original_serial_loss[0].name]
            else:
                self._serial_loss = self._original_serial_loss
            self._serial_optimizer = self._original_serial_optimizer
            self._init_dist_attr_for_program()
            self._tensors_ids = list(self._dist_tensors_for_program.keys())
            self._ops_ids = list(self._dist_ops_for_program.keys())
            set_flags({"FLAGS_convert_all_blocks": True})
            self._serial_graph = framework.IrGraph(
                core.Graph(self._serial_main_program.desc))
            self._init_dist_attr_for_graph()
            self._is_initialized = True

    # def reset(self,
    #           skip_dist_tensors=None,
    #           skip_dist_ops=None,
    #           skip_tensor_dist_attr_fields=None,
    #           skip_op_dist_attr_fields=None):
    #     self._serial_main_program = self._original_serial_main_program.clone()
    #     self._serial_startup_program = self._original_serial_startup_program.clone()
    #     new_tensors_ids = []
    #     for tensor_id, dist_tensor in self._dist_tensors_for_program.items():
    #         if tensor_id in self._tensors_ids:
    #             dist_tensor.dist_attr.reset(skip_tensor_dist_attr_fields)
    #         else:
    #             new_tensors_ids.append(tensor_id)
    #     for tensor_id in new_tensors_ids:
    #         self._dist_tensors_for_program.pop(tensor_id)
    #     new_ops_ids = []
    #     for op_id, dist_op in self._dist_ops_for_program.items():
    #         if op_id in self._ops_ids:
    #             dist_op.dist_attr.reset(skip_op_dist_attr_fields)
    #         else:
    #             new_ops_ids.append(op_id)
    #     for op_id in new_ops_ids:
    #         self._dist_ops_for_program.pop(op_id)

    #     self.copy_dist_attr_from_program_to_graph()

    #     self._dist_main_programs = {}
    #     self._dist_startup_programs = {}

    #     self._pass_context = PassContext()

    #     self._dist_op_context = DistributedOperatorContext()

    #     self._process_meshes = []

    def add_process_mesh(self, process_mesh):
        assert isinstance(process_mesh, ProcessMesh), \
            'The type of dim_mapping must be ProcessMesh.'
        if process_mesh not in self.process_meshes:
            self._process_meshes.append(process_mesh)

    def add_dist_tensor_for_program(self, dist_tensor):
        inner_serial_tensor = dist_tensor.serial_tensor
        inner_serial_tensor_id = inner_serial_tensor.desc.original_id()
        self._dist_tensors_for_program[inner_serial_tensor_id] = dist_tensor

    def add_dist_op_for_program(self, dist_op):
        inner_serial_op = dist_op.serial_op
        inner_serial_op_id = inner_serial_op.desc.original_id()
        self._dist_ops_for_program[inner_serial_op_id] = dist_op

    def get_dist_tensor_for_program(self, serial_tensor):
        serial_tensor_id = serial_tensor.desc.id()
        dist_tensor = self._dist_tensors_for_program.get(serial_tensor_id, None)
        if dist_tensor:
            return dist_tensor
        else:
            serial_tensor_id = serial_tensor.desc.original_id()
            dist_tensor = self._dist_tensors_for_program.get(serial_tensor_id,
                                                             None)
            if dist_tensor:
                return dist_tensor
            else:
                return None

    def get_dist_tensor_for_graph(self, serial_tensor_node):
        serial_tensor_node_id = _node_id(serial_tensor_node)
        return self._dist_tensors_for_graph.get(serial_tensor_node_id, None)

    def get_dist_op_for_program(self, serial_op):
        serial_op_id = serial_op.desc.id()
        dist_op = self._dist_ops_for_program.get(serial_op_id, None)
        if dist_op:
            return dist_op
        else:
            serial_op_id = serial_op.desc.original_id()
            dist_op = self._dist_ops_for_program.get(serial_op_id, None)
            if dist_op:
                return dist_op
            else:
                return None

    def del_dist_op_for_program(self, serial_tensor):
        serial_tensor_id = serial_tensor.desc.id()
        if self._dist_ops_for_program.get(serial_tensor_id, None):
            del self._dist_ops_for_program[serial_tensor_id]

    def get_dist_op_for_graph(self, serial_op_node):
        serial_op_node_id = _node_id(serial_op_node)
        return self._dist_ops_for_graph.get(serial_op_node_id, None)

    def get_tensor_dist_attr_for_program(self, serial_tensor):
        serial_tensor_id = serial_tensor.desc.id()
        dist_tensor = self._dist_tensors_for_program.get(serial_tensor_id, None)
        if dist_tensor:
            return dist_tensor.dist_attr
        else:
            serial_tensor_id = serial_tensor.desc.original_id()
            dist_tensor = self._dist_tensors_for_program.get(serial_tensor_id,
                                                             None)
            if dist_tensor:
                return dist_tensor.dist_attr
            else:
                return None

    def get_tensor_dist_attr_for_program_with_id(self, tensor_id):
        dist_tensor = self._dist_tensors_for_program.get(tensor_id, None)
        if dist_tensor:
            return dist_tensor.dist_attr
        else:
            return None

    def set_tensor_dist_attr_for_program(self, serial_tensor, dist_attr):
        dist_tensor = DistributedTensor(serial_tensor, dist_attr)
        self.add_dist_tensor_for_program(dist_tensor)

    def get_tensor_dist_attr_for_graph(self, serial_tensor_node):
        serial_tensor_node_id = _node_id(serial_tensor_node)
        dist_tensor = self._dist_tensors_for_graph.get(serial_tensor_node_id,
                                                       None)
        if dist_tensor:
            return dist_tensor.dist_attr
        else:
            return None

    def get_op_dist_attr_for_program(self, serial_op):
        serial_op_id = serial_op.desc.id()
        dist_op = self._dist_ops_for_program.get(serial_op_id, None)
        if dist_op:
            return dist_op.dist_attr
        else:
            serial_op_id = serial_op.desc.original_id()
            dist_op = self._dist_ops_for_program.get(serial_op_id, None)
            if dist_op:
                return dist_op.dist_attr
            else:
                return None

    def get_op_dist_attr_for_program_with_id(self, op_id):
        dist_op = self._dist_ops_for_program.get(op_id, None)
        if dist_op:
            return dist_op.dist_attr
        else:
            return None

    def set_op_dist_attr_for_program(self, serial_op, dist_attr):
        dist_op = DistributedOperator(serial_op, dist_attr)
        self.add_dist_op_for_program(dist_op)

    def get_op_dist_attr_for_graph(self, serial_op_node):
        serial_op_node_id = _node_id(serial_op_node)
        dist_op = self._dist_ops_for_graph.get(serial_op_node_id, None)
        if dist_op:
            return dist_op.dist_attr
        else:
            return None

    def get_dist_attr_for_graph(self, serial_node):
        if serial_node.is_var() and serial_node.var() is not None:
            serial_tensor_node_id = _node_id(serial_node)
            dist_tensor = self._dist_tensors_for_graph.get(
                serial_tensor_node_id, None)
            if dist_tensor:
                return dist_tensor.dist_attr
            else:
                return None
        if serial_node.is_op() and serial_node.op() is not None:
            serial_op_node_id = _node_id(serial_node)
            dist_op = self._dist_ops_for_graph.get(serial_op_node_id, None)
            if dist_op:
                return dist_op.dist_attr
            else:
                return None
        return None

    def _init_dist_attr_for_program(self, no_default=False):
        # Copy the dist tensors and dist ops annotated by users from the default context
        if not no_default:
            default_ctx = get_default_distributed_context()
            self._process_meshes = copy.deepcopy(default_ctx.process_meshes)
        else:
            default_ctx = self
        for block in self._serial_main_program.blocks:
            for tensor in block.vars.values():
                # Copy the distributed tensors in the default context
                default_dist_tensor = default_ctx.get_dist_tensor_for_program(
                    tensor)
                if default_dist_tensor and default_ctx is not self:
                    self.add_dist_tensor_for_program(default_dist_tensor)
                current_dist_tensor = self.get_dist_tensor_for_program(tensor)
                if current_dist_tensor is None:
                    dist_tensor = DistributedTensor(tensor)
                    self.add_dist_tensor_for_program(dist_tensor)
            for op in block.ops:
                # Copy the distributed operators in the default context
                default_dist_op = default_ctx.get_dist_op_for_program(op)
                if default_dist_op and default_ctx is not self:
                    self.add_dist_op_for_program(default_dist_op)
                current_dist_op = self.get_dist_op_for_program(op)
                if current_dist_op is None:
                    dist_op = DistributedOperator(op)
                    self.add_dist_op_for_program(dist_op)

    def _order_nodes_by_program_order(self):
        def _contains(nodes, target_node):
            for node in nodes:
                if _node_id(node) == _node_id(target_node):
                    return True
            return False

        serial_ordered_tensor_nodes = []
        serial_ordered_op_nodes = []
        all_nodes = []
        for idx, graph in enumerate(self._serial_graph.all_sub_graphs()):
            for node in graph.all_nodes():
                all_nodes.append(node)
        for node in all_nodes:
            if node.is_var() and node.var() is not None:
                serial_ordered_tensor_nodes.append(node)
            if node.is_op() and node.op() is not None:
                serial_ordered_op_nodes.append(node)
        serial_ordered_tensor_nodes.sort(
            key=lambda node: node.node.original_desc_id())
        serial_ordered_op_nodes.sort(
            key=lambda node: node.node.original_desc_id())
        num_nodes_before = len(serial_ordered_tensor_nodes) + len(
            serial_ordered_op_nodes)

        new_serial_ordered_tensor_nodes = []
        new_serial_ordered_op_nodes = []
        new_serial_ordered_nodes = []
        for op_node in serial_ordered_op_nodes:
            tensor_nodes = []
            for tensor_node in op_node.inputs:
                if tensor_node.is_var() \
                    and tensor_node.var() is not None \
                    and not _contains(new_serial_ordered_nodes, tensor_node):
                    tensor_nodes.append(tensor_node)
                    new_serial_ordered_tensor_nodes.append(tensor_node)
            tensor_nodes.sort(key=lambda node: node.node.original_desc_id())
            new_serial_ordered_nodes.extend(tensor_nodes)
            new_serial_ordered_nodes.append(op_node)
            new_serial_ordered_op_nodes.append(op_node)
            tensor_nodes = []
            for tensor_node in op_node.outputs:
                if tensor_node.is_var() \
                    and tensor_node.var() is not None \
                    and not _contains(new_serial_ordered_nodes, tensor_node):
                    tensor_nodes.append(tensor_node)
                    new_serial_ordered_tensor_nodes.append(tensor_node)
            tensor_nodes.sort(key=lambda node: node.node.original_desc_id())
            new_serial_ordered_nodes.extend(tensor_nodes)
        new_serial_ordered_tensor_nodes.sort(
            key=lambda node: node.node.original_desc_id())
        new_serial_ordered_op_nodes.sort(
            key=lambda node: node.node.original_desc_id())
        self._serial_ordered_tensor_nodes = new_serial_ordered_tensor_nodes
        self._serial_ordered_op_nodes = new_serial_ordered_op_nodes
        self._serial_ordered_nodes = new_serial_ordered_nodes
        assert len(self._serial_ordered_nodes) == len(
            self._serial_ordered_tensor_nodes) + len(
                self._serial_ordered_op_nodes)
        self._serial_orphan_tensor_nodes = []
        for tensor_node in serial_ordered_tensor_nodes:
            if not _contains(self._serial_ordered_tensor_nodes, tensor_node):
                self._serial_orphan_tensor_nodes.append(tensor_node)
        if len(self._serial_ordered_nodes) != num_nodes_before:
            print(
                "WARNING: there are some orphan tensors or ops which are not used in the execution."
            )

    def _init_dist_attr_for_graph(self):
        # Convert program to graph and initialize the distributed attributes
        self._order_nodes_by_program_order()
        for node in self.serial_ordered_nodes:
            if node.is_var() and node.var() is not None:
                dist_tensor = None
                tensor_id = node.node.original_desc_id()
                for cur_tensor_id, cur_dist_tensor in self._dist_tensors_for_program.items(
                ):
                    if tensor_id == cur_tensor_id \
                        or tensor_id == cur_dist_tensor.serial_tensor.desc.original_id():
                        dist_tensor = cur_dist_tensor
                        self._node_id_to_tensor_id[_node_id(
                            node)] = cur_tensor_id
                assert dist_tensor is not None, \
                    "Tensor must have a distributed tensor after the initialization for program."
                serial_tensor_node_id = _node_id(node)
                new_dist_tensor = DistributedTensor(dist_tensor.serial_tensor,
                                                    dist_tensor.dist_attr)
                self._dist_tensors_for_graph[
                    serial_tensor_node_id] = new_dist_tensor
            if node.is_op() and node.op() is not None:
                dist_op = None
                op_id = node.node.original_desc_id()
                for cur_op_id, cur_dist_op in self._dist_ops_for_program.items(
                ):
                    if op_id == cur_op_id \
                        or op_id == cur_dist_op.serial_op.desc.original_id():
                        dist_op = cur_dist_op
                        self._node_id_to_op_id[_node_id(node)] = cur_op_id
                assert dist_op is not None, \
                    "Operator must have a distributed operator after the initialization for program."
                serial_op_node_id = _node_id(node)
                new_dist_op = DistributedOperator(dist_op.serial_op,
                                                  dist_op.dist_attr)
                self._dist_ops_for_graph[serial_op_node_id] = new_dist_op

    def clear_dist_info_for_program(self):
        self._dist_tensors_for_program.clear()
        self._dist_ops_for_program.clear()

    def clear_dist_info_for_graph(self):
        self._dist_tensors_for_graph.clear()
        self._dist_ops_for_graph.clear()

    def copy_dist_attr_from_program_to_graph(self):
        for node in self.serial_ordered_nodes:
            if node.is_var() and node.var() is not None:
                dist_tensor = None
                tensor_id = node.node.original_desc_id()
                for cur_tensor_id, cur_dist_tensor in self._dist_tensors_for_program.items(
                ):
                    if tensor_id == cur_tensor_id \
                        or tensor_id == cur_dist_tensor.serial_tensor.desc.original_id():
                        dist_tensor = cur_dist_tensor
                assert dist_tensor is not None, \
                    "Tensor must have a distributed tensor after the initialization for program."
                serial_tensor_node_id = _node_id(node)
                new_dist_tensor = DistributedTensor(dist_tensor.serial_tensor,
                                                    dist_tensor.dist_attr)
                self._dist_tensors_for_graph[
                    serial_tensor_node_id] = new_dist_tensor
            if node.is_op() and node.op() is not None:
                dist_op = None
                op_id = node.node.original_desc_id()
                for cur_op_id, cur_dist_op in self._dist_ops_for_program.items(
                ):
                    if op_id == cur_op_id \
                        or op_id == cur_dist_op.serial_op.desc.original_id():
                        dist_op = cur_dist_op
                assert dist_op is not None, \
                    "Operator must have a distributed operator after the initialization for program."
                serial_op_node_id = _node_id(node)
                new_dist_op = DistributedOperator(dist_op.serial_op,
                                                  dist_op.dist_attr)
                self._dist_ops_for_graph[serial_op_node_id] = new_dist_op

    def copy_dist_attr_from_graph_to_program(self):
        assert self._is_initialized, \
            "Both program and graph must be initialized."
        updated_tensors = {}
        # all_nodes = self._serial_graph.all_nodes()
        all_nodes = self._serial_ordered_nodes
        for node in all_nodes:
            if node.is_var() and node.var() is not None:
                tensor_id = self._node_id_to_tensor_id[_node_id(node)]
                updated = updated_tensors.get(tensor_id, False)
                # If a var has multiples var nodes in graph, only use the first one for now
                if not updated:
                    tensor_dist_attr_for_graph = self.get_tensor_dist_attr_for_graph(
                        node)
                    dist_tensor_for_program = self._dist_tensors_for_program[
                        tensor_id]
                    dist_tensor_for_program.dist_attr = tensor_dist_attr_for_graph
                    updated_tensors[tensor_id] = True
            if node.is_op() and node.op() is not None:
                op_id = self._node_id_to_op_id[_node_id(node)]
                op_dist_attr_for_graph = self.get_op_dist_attr_for_graph(node)
                dist_op_for_program = self._dist_ops_for_program[op_id]
                dist_op_for_program.dist_attr = op_dist_attr_for_graph
        # TODO: the completion algorithm will skip orphan tensors,
        # here we just set there process_mesh to the first one.
        for orphan_node in self._serial_orphan_tensor_nodes:
            serial_tensor_id = orphan_node.var().id()
            dist_tensor = self._dist_tensors_for_program.get(serial_tensor_id,
                                                             None)
            if dist_tensor:
                dist_tensor.dist_attr.process_mesh = self._process_meshes[0]
            else:
                serial_tensor_id = orphan_node.var().original_id()
                dist_tensor = self._dist_tensors_for_program.get(
                    serial_tensor_id, None)
                dist_tensor.dist_attr.process_mesh = self._process_meshes[0]

    def amend_dist_attr_for_program(self):
        for dist_tensor in self._dist_tensors_for_program.values():
            serial_tensor = dist_tensor.serial_tensor
            dist_attr = dist_tensor.dist_attr
            if serial_tensor.type == core.VarDesc.VarType.READER \
                or serial_tensor.type == core.VarDesc.VarType.LOD_TENSOR_ARRAY \
                or serial_tensor.type == core.VarDesc.VarType.STEP_SCOPES:
                tensor_shape = []
            else:
                tensor_shape = serial_tensor.shape
            dims_mapping = dist_attr.dims_mapping
            process_mesh_shape = dist_attr.process_mesh.topology
            # If the dimension of tensor is less than the sharding dimension of process mesh,
            # we just amend the dimension mapping to -1. (Is this really OK?)
            for i in range(len(tensor_shape)):
                if dims_mapping[i] != -1 and tensor_shape[i] > 0 \
                    and process_mesh_shape[dims_mapping[i]] > tensor_shape[i]:
                    dims_mapping[i] = -1

        for dist_op in self._dist_ops_for_program.values():
            serial_op = dist_op.serial_op
            dist_attr = dist_op.dist_attr
            for arg_name in serial_op.input_arg_names:
                if dist_op.get_serial_input(arg_name) is None:
                    tensor_shape = []
                else:
                    if dist_op.get_serial_input(arg_name).type == core.VarDesc.VarType.READER \
                        or dist_op.get_serial_input(arg_name).type == core.VarDesc.VarType.LOD_TENSOR_ARRAY \
                        or dist_op.serial_op.type == "create_py_reader":
                        tensor_shape = []
                    else:
                        tensor_shape = dist_op.get_serial_input(arg_name).shape
                dims_mapping = dist_attr.get_input_dims_mapping(arg_name)
                process_mesh_shape = dist_attr.process_mesh.topology
                # If the dimension of tensor is less than the sharding dimension of process mesh,
                # we just amend the dimension mapping to -1. (Is this really OK?)
                for i in range(len(tensor_shape)):
                    if dims_mapping[i] != -1 and tensor_shape[i] > 0 \
                        and process_mesh_shape[dims_mapping[i]] > tensor_shape[i]:
                        dims_mapping[i] = -1
            for arg_name in serial_op.output_arg_names:
                if dist_op.get_serial_output(arg_name).type == core.VarDesc.VarType.READER \
                    or dist_op.get_serial_output(arg_name).type == core.VarDesc.VarType.LOD_TENSOR_ARRAY \
                    or dist_op.get_serial_output(arg_name).type == core.VarDesc.VarType.STEP_SCOPES:
                    tensor_shape = []
                else:
                    tensor_shape = dist_op.get_serial_output(arg_name).shape
                dims_mapping = dist_attr.get_output_dims_mapping(arg_name)
                process_mesh_shape = dist_attr.process_mesh.topology
                # If the dimension of tensor is less than the sharding dimension of process mesh,
                # we just amend the dimension mapping to -1. (Is this really OK?)
                for i in range(len(tensor_shape)):
                    if dims_mapping[i] != -1 and tensor_shape[i] > 0 \
                        and process_mesh_shape[dims_mapping[i]] > tensor_shape[i]:
                        dims_mapping[i] = -1

    def validate_dist_attr_for_program(self):
        if not self._is_initialized:
            assert False, \
                "Program must be initialized before validating its distributed attributes"
        for block in self.serial_main_program.blocks:
            for tensor in block.vars.values():
                dist_tensor = self.get_dist_tensor_for_program(tensor)
                assert dist_tensor is not None, \
                    "Tensor {} does not have a distributed attribute.".format(
                        dist_tensor.serial_tensor.name)
                if (dist_tensor is not None) and (
                        not dist_tensor.validate_dist_attr()):
                    assert False, "Tensor {} has a wrong distributed attributes {}.".format(
                        dist_tensor.serial_tensor.name, dist_tensor.dist_attr)
            for op in block.ops:
                dist_op = self.get_dist_op_for_program(op)
                assert dist_op is not None, \
                    "Operator {} does not have a distributed attribute.".format(
                        dist_op.serial_op.type)
                if (dist_op is not None) and (not dist_op.validate_dist_attr()):
                    assert False, "Operator {} has a wrong distributed attributes {}.".format(
                        dist_op.serial_op.type, dist_op.dist_attr)
        return True

    def __deepcopy__(self, memo):
        cls = self.__class__
        result = cls.__new__(cls)
        memo[id(self)] = result
        for k, v in self.__dict__.items():
            if k in [
                "_original_serial_main_program", "_original_serial_startup_program", \
                "_serial_main_program", "_serial_startup_program", "_serial_graph", \
                "_dist_main_programs", "_dist_startup_programs", \
                "_serial_ordered_nodes", "_serial_ordered_tensor_nodes", \
                "_serial_ordered_op_nodes"]:
                setattr(result, k, v)
            else:
                setattr(result, k, copy.deepcopy(v, memo))

        # update dist tensor's dist_context
        for key in result._dist_tensors_for_program.keys():
            result._dist_tensors_for_program[key]._dist_context = result
        return result


class DistributedOperatorContext:
    """
    DistributedOperatorContext is used to create a dist op desc in Program.
    Every time to create a new dist op, the context should be updated for it accordingly.
    """

    def __init__(self):
        self._dst_main_program = None
        self._main_block = None
        self._dst_startup_program = None
        self._startup_block = None
        self._cur_src_op = None
        self._cur_dist_attr = None
        self.grad_op_id_to_op_id = {}
        self.grad_var_to_var = defaultdict(dict)
        self._work_block = None
        self.already_init_sync_vars = set()
        self.varname_mapping = None
        self.rank_id = None

    def __deepcopy__(self, memo):
        cls = self.__class__
        result = cls.__new__(cls)
        memo[id(self)] = result
        for k, v in self.__dict__.items():
            if k in [
                    "_dst_main_program", "_dst_startup_program", "_cur_src_op",
                    "_work_block", "_main_block", "_startup_block"
            ]:
                setattr(result, k, v)
            else:
                setattr(result, k, copy.deepcopy(v, memo))
        return result

    @property
    def dst_main_program(self):
        return self._dst_main_program

    @dst_main_program.setter
    def dst_main_program(self, prog):
        self._dst_main_program = prog
        self._main_block = prog.blocks[0]

    @property
    def main_block(self):
        return self._main_block

    @property
    def dst_startup_program(self):
        return self._dst_startup_program

    @dst_startup_program.setter
    def dst_startup_program(self, prog):
        self._dst_startup_program = prog
        self._startup_block = prog.blocks[0]

    @property
    def startup_block(self):
        return self._startup_block

    @property
    def work_block(self):
        assert self._work_block is not None
        return self._work_block

    @work_block.setter
    def work_block(self, block):
        assert block is not None
        self._work_block = block

    @property
    def cur_src_op(self):
        assert self._cur_src_op is not None
        return self._cur_src_op

    def prepare_context(self, src_op):

        self._cur_src_op = src_op

        # build input varname mapping
        kinputs = {}
        for input_name in src_op.desc.input_names():
            varnames = []
            for varname in src_op.desc.input(input_name):
                assert varname in self.varname_mapping
                varnames.append(self.varname_mapping[varname])
            kinputs[input_name] = varnames

        # build output varname mapping
        koutputs = {}
        for output_name in src_op.desc.output_names():
            varnames = []
            for varname in src_op.desc.output(output_name):
                assert varname in self.varname_mapping
                varnames.append(self.varname_mapping[varname])
            koutputs[output_name] = varnames

        return kinputs, koutputs


class BlockState(object):
    def __init__(self):
        self.nblock = 0
        self.forward_indices = []
        self.backward_indices = []
        self.backward_to_forward_index_map = {}

    def parse_forward_blocks(self, program):

        while program.current_block_idx != 0:
            program._rollback()

        assert program.current_block_idx == 0

        for idx, block in enumerate(program.blocks):

            assert idx == block.idx, "index doesn't match"
            assert block.forward_block_idx == -1, "forward_block_idx of forward block [{}] is not [{}]".format(
                idx, block.forward_block_idx)
            self.forward_indices.append(idx)
            self.nblock += 1

        assert self.nblock >= 1

    def parse_backward_blocks(self, program):

        assert 0 in self.forward_indices, "forward block idx are{}".format(
            self.forward_indices)
        self.backward_to_forward_index_map[0] = 0

        for idx, block in enumerate(program.blocks):

            if idx < len(self.forward_indices):
                continue

            assert idx == block.idx, "index doesn't match"
            assert block.forward_block_idx in self.forward_indices
            self.backward_indices.append(idx)
            self.backward_to_forward_index_map[idx] = block.forward_block_idx
            self.nblock += 1

        assert self.nblock == len(program.blocks)