Merge pull request #2258 from hpcaitech/debug/ckpt-autoparallel

[autockpt] provide option for activation checkpoint search in SPMD solver

colossalai/auto_parallel/checkpoint/ckpt_solver_base.py

@@ -5,8 +5,12 @@ from typing import Any, List
 import torch
 from torch.fx import Graph, Node
 
+from colossalai.auto_parallel.passes.runtime_apply_pass import (
+    runtime_apply,
+    runtime_apply_for_iterable_object,
+    runtime_comm_spec_apply,
+)
 from colossalai.fx.codegen.activation_checkpoint_codegen import ActivationCheckpointCodeGen
-from colossalai.fx.profiler.memory_utils import is_inplace
 
 __all___ = ['CheckpointSolverBase']
 
@@ -31,10 +35,11 @@ class CheckpointSolverBase(ABC):
         free_memory: float = -1.0,
         requires_linearize: bool = False,
         cnode: List[str] = None,
+        optim_multiplier: float = 1.0,
     ):
-        """CheckpointSolver class will integrate information provided by the components
+        """``CheckpointSolverBase`` class will integrate information provided by the components
-        and use an existing solver to find a possible optimal strategies combination for
+        and use an existing solver to find a possible optimal strategies combination for target
-        target computing graph.
+        computing graph.
 
         Existing Solvers:
             Chen's Greedy solver: https://arxiv.org/abs/1604.06174  (CheckpointSolverChen)
@@ -45,9 +50,11 @@ class CheckpointSolverBase(ABC):
             free_memory (float): Memory constraint for the solution.
             requires_linearize (bool): Whether the graph needs to be linearized.
             cnode (List[str], optional): Common node List, should be the subset of input. Default to None.
+            optim_multiplier (float, optional): The multiplier of extra weight storage for the
+            ``torch.optim.Optimizer``. Default to 1.0.
 
         Warnings:
-            `MetaInfoProp` should be done before constructing the solver. Meta information of the graph is required.
+            Meta information of the graph is required for any ``CheckpointSolver``.
         """
         # super-dainiu: this graph is a temporary graph which can refer to
         # the owning module, but we will return another deepcopy of it after
@@ -57,13 +64,14 @@ class CheckpointSolverBase(ABC):
         _copy_output(graph, self.graph)
         self.graph.set_codegen(ActivationCheckpointCodeGen())
 
-        # check if `MetaInfoProp` is done
+        # check if has meta information
         if any(len(node.meta) == 0 for node in self.graph.nodes):
             raise RuntimeError(
-                "Nodes meta information hasn't been prepared! Please run MetaInfoProp before constructing the solver!")
+                "Nodes meta information hasn't been prepared! Please extract from graph before constructing the solver!"
+            )
 
-        self.free_memory = free_memory
+        # parameter memory = parameter size + optimizer extra weight storage
-        self.parameter_size = _get_param_size(self.graph.owning_module)
+        self.free_memory = free_memory - _get_param_size(self.graph.owning_module) * (optim_multiplier + 1)
         self.cnode = cnode
         self.requires_linearize = requires_linearize
         if self.requires_linearize:
@@ -93,7 +101,7 @@ class CheckpointSolverBase(ABC):
             the actual 'node' in linearized manner.
 
         Remarks:
-            Do merge the inplace ops into the previous node.
+            Do merge the inplace ops and shape-consistency ops into the previous node.
         """
 
         # Common nodes are type of nodes that could be seen as attributes and remain
@@ -131,7 +139,23 @@ class CheckpointSolverBase(ABC):
                 bool
             """
 
-            return not sum([v for _, v in deps.items()]) and not any(map(is_inplace, n.users))
+            def _is_inplace(n: Node):
+                """Get the inplace argument from ``torch.fx.Node``
+                """
+                inplace = False
+                if n.op == "call_function":
+                    inplace = n.kwargs.get("inplace", False)
+                elif n.op == "call_module":
+                    inplace = getattr(n.graph.owning_module.get_submodule(n.target), "inplace", False)
+                return inplace
+
+            def _is_shape_consistency(n: Node):
+                """Check if this node is shape-consistency node (i.e. ``runtime_apply`` or ``runtime_apply_for_iterable_object``)
+                """
+                return n.target in [runtime_apply, runtime_apply_for_iterable_object, runtime_comm_spec_apply]
+
+            return not sum([v for _, v in deps.items()]) and not any(map(_is_inplace, n.users)) and not any(
+                map(_is_shape_consistency, n.users))
 
         # make sure that item in cnode is valid
         if self.cnode:

colossalai/auto_parallel/checkpoint/ckpt_solver_chen.py

@@ -19,9 +19,9 @@ class CheckpointSolverChen(CheckpointSolverBase):
         Note that this algorithm targets at memory optimization only, using techniques in appendix A.
 
         Usage:
-            Assume that we have a `GraphModule`, and we already applied the `MetaInfoProp`
+            Assume that we have a ``GraphModule``, and we have already done the extractions
             to the graph to retrieve all information needed, then we could use the following
-            code to find a solution using `CheckpointSolverChen`:
+            code to find a solution using ``CheckpointSolverChen``:
             >>> solver = CheckpointSolverChen(gm.graph)
             >>> chen_graph = solver.solve()
             >>> gm.graph = chen_graph    # set the graph to a new graph
@@ -74,7 +74,7 @@ class CheckpointSolverChen(CheckpointSolverBase):
     def grid_search(self) -> Set:
         """
         Search ckpt strategy with b = 0, then run the allocation algorithm again with b = √xy.
-        Grid search over [√2/2 b, √2 b] for ckpt_opt over num_grids as in appendix A.
+        Grid search over [√2/2 b, √2 b] for ``ckpt_opt`` over ``num_grids`` as in appendix A.
         """
         _, b_approx = self.run_chen_greedy(0)
         b_min, b_max = math.floor(b_approx / math.sqrt(2)), math.ceil(b_approx * math.sqrt(2))

colossalai/auto_parallel/checkpoint/ckpt_solver_rotor.py

@@ -4,6 +4,7 @@ from typing import Any, Dict, List, Tuple
 from torch import Tensor
 from torch.fx import Graph, Node
 
+from colossalai.auto_parallel.passes.runtime_apply_pass import runtime_apply, runtime_comm_spec_apply
 from colossalai.fx.codegen.activation_checkpoint_codegen import _find_nested_ckpt_regions
 from colossalai.fx.profiler import (
     activation_size,
@@ -22,15 +23,20 @@ __all__ = ['CheckpointSolverRotor']
 
 class CheckpointSolverRotor(CheckpointSolverBase):
 
-    def __init__(self, graph: Graph, free_memory: float = -1, cnode: List[str] = None, memory_slots: int = 500):
+    def __init__(self,
+                 graph: Graph,
+                 free_memory: float = -1,
+                 cnode: List[str] = None,
+                 memory_slots: int = 500,
+                 optim_multiplier: float = 1.0):
         """This is the simple implementation of dynamic programming algorithm rotor
         in https://hal.inria.fr/hal-02352969. Some code are adapted from
         https://gitlab.inria.fr/hiepacs/rotor.
 
         Usage:
-            Assume that we have a `GraphModule`, and we already applied the `MetaInfoProp`
+            Assume that we have a ``GraphModule``, and we have already done the extractions
             to the graph to retrieve all information needed, then we could use the following
-            code to find a solution using `CheckpointSolverRotor`:
+            code to find a solution using ``CheckpointSolverRotor``:
             >>> solver = CheckpointSolverRotor(gm.graph, free_memory=torch.cuda.mem_get_info(device=0)[0])
             >>> rotor_graph = solver.solve(force_python=True)   # otherwise use C solver
             >>> gm.graph = rotor_graph    # set the graph to a new graph
@@ -41,8 +47,10 @@ class CheckpointSolverRotor(CheckpointSolverBase):
                 Use ``torch.cuda.mem_get_info(device=0)[0]`` to estimate the free_memory. Defaults to -1.
             cnode (List[str], optional): Common node List, should be the subset of input. Defaults to None.
             memory_slots (int, optional): Number of slots for discretizing memory budget. Defaults to 500.
+            optim_multiplier (float, optional): The multiplier of extra weight storage for the
+            ``torch.optim.Optimizer``. Default to 1.0.
         """
-        super().__init__(graph, free_memory, True, cnode)
+        super().__init__(graph, free_memory, True, cnode, optim_multiplier)
         self.memory_slots = memory_slots
 
         # construct chain
@@ -128,16 +136,24 @@ class CheckpointSolverRotor(CheckpointSolverBase):
         xbar = 0
         ftime = 0
         btime = 0
+        fwd_mem_peak = 0
         for n in node:
             assert isinstance(n, Node), f'{n} is not a Node'
-            xbar += calculate_fwd_tmp(n) + calculate_fwd_out(n)
+            if n.target == runtime_apply or n.target == runtime_comm_spec_apply:
+                # in this case we need to calculate memory usage directly based on the statics that hooked in node.meta
+                xbar += n.meta['fwd_mem_out']
+                fwd_mem_peak = max(fwd_mem_peak, xbar + n.meta['fwd_mem_tmp'])
+            else:
+                xbar += calculate_fwd_tmp(n) + calculate_fwd_out(n)
+                fwd_mem_peak = max(fwd_mem_peak, xbar + n.meta['fwd_mem_tmp'] + cls._extract_unused_output(n))
+
             # minimum flop count is required
             ftime += max(calculate_fwd_time(n), 1.0)
             btime += max(calculate_bwd_time(n), 1.0)
 
         x = calculate_fwd_out(node[-1])
         xbar = max(x, xbar)
-        ftmp = cls._extract_ftmp(node)
+        ftmp = fwd_mem_peak - xbar
         btmp = cls._extract_btmp(node)
         return ftime, btime, x, xbar, ftmp, btmp
 
@@ -151,10 +167,9 @@ class CheckpointSolverRotor(CheckpointSolverBase):
         return input_tensors
 
     @staticmethod
-    def _extract_ftmp(node: List[Node]) -> int:
+    def _extract_unused_output(node: Node) -> int:
-        """Extract ftmp from a list of nodes"""
+        """Extract unused output from `torch.fx.Node`"""
-        n = node[-1]
+        return activation_size(node.meta['fwd_out']) - calculate_fwd_out(node)
-        return activation_size(n.meta['fwd_out']) - calculate_fwd_out(n)
 
     @staticmethod
     def _extract_btmp(node: List[Node]) -> int:
@@ -290,8 +305,8 @@ class CheckpointSolverRotor(CheckpointSolverBase):
             lhs (int): The left index of the interval to backtrack.
             rhs (int): The right index of the interval to backtrack.
             budget (int): The memory budget for processing this interval.
-            cost_table (List[Any]): See `._compute_table()` for definitions
+            cost_table (List[Any]): See ``._compute_table()`` for definitions
-            back_ptr (List[Any]): See `._compute_table()` for definitions
+            back_ptr (List[Any]): See ``._compute_table()`` for definitions
 
         Raises:
             ValueError: Can not process the chain.
@@ -332,7 +347,7 @@ class CheckpointSolverRotor(CheckpointSolverBase):
 
     @staticmethod
     def _annotate_from_sequence(sequence: Sequence, node_list: List[List[Node]]):
-        """Annotate the nodes in the node_list with activation checkpoint from the sequence.
+        """Annotate the nodes in the ``node_list`` with activation checkpoint from the sequence.
 
         Args:
             sequence (Sequence): The sequence of executing nodes with activation checkpoint annotations.

colossalai/auto_parallel/meta_profiler/constants.py

@@ -5,8 +5,11 @@ import torch.nn as nn
 
 from ..tensor_shard.constants import *
 
-# list of inplace operations
+# list of inplace module
 INPLACE_MODULE = [nn.ReLU]
 
+# list of inplace operations
+INPLACE_OPS = [torch.flatten]
+
 # list of operations that do not save forward activations
 NO_SAVE_ACTIVATION = [torch.add, torch.sub, operator.add, operator.sub]

colossalai/auto_parallel/meta_profiler/meta_registry/binary_elementwise_ops.py

@@ -24,26 +24,25 @@ def binary_elementwise_meta_info(*args, **kwargs) -> Tuple[TrainCycleItem, Train
         Tuple[TrainCycleItem, TrainCycleItem, List[torch.Tensor]]: compute cost, memory cost and forward inputs
     """
 
-    input_op_data, other_op_data = [arg for arg in args if arg.type != OperationDataType.OUTPUT]
+    input_op_data = [arg for arg in args if arg.type != OperationDataType.OUTPUT]
     output_op_data = next(filter(lambda arg: arg.type == OperationDataType.OUTPUT, args))
 
     # construct forward args for flop mapping
-    fwd_in_args = [input_op_data.data, other_op_data.data]
+    fwd_in_args = [opdata.data for opdata in input_op_data]
     fwd_out_args = [output_op_data.data]
 
     # calculate cost
 
     # calculate compute cost
     # NOTE: we set bwd_compute_cost two times of fwd_compute_cost in this case
-    fwd_compute_cost = flop_mapping[torch.ops.aten._adaptive_avg_pool2d.default](fwd_in_args, fwd_out_args)
+    fwd_compute_cost = flop_mapping[torch.ops.aten.add.Tensor](fwd_in_args, fwd_out_args)
     bwd_compute_cost = fwd_compute_cost * 2
     compute_cost = TrainCycleItem(fwd=fwd_compute_cost, bwd=bwd_compute_cost, total=fwd_compute_cost + bwd_compute_cost)
 
     # calculate memory cost
-    param_mem_cost = activation_size(
+    param_mem_cost = activation_size([arg.data for arg in input_op_data if arg.type == OperationDataType.PARAM])
-        [arg.data for arg in [input_op_data, other_op_data] if arg.type == OperationDataType.PARAM])
     fwd_mem_cost = MemoryCost(
-        activation=activation_size([input_op_data.data, output_op_data.data]),
+        activation=activation_size(output_op_data.data),
         parameter=param_mem_cost,
     )
     bwd_mem_cost = MemoryCost(
@@ -60,7 +59,7 @@ def binary_elementwise_meta_info(*args, **kwargs) -> Tuple[TrainCycleItem, Train
     memory_cost = TrainCycleItem(fwd=fwd_mem_cost, bwd=bwd_mem_cost, total=total_mem_cost)
 
     # store fwd_in, fwd_buffer, fwd_out
-    fwd_in = [torch.zeros_like(input_op_data.data, device='meta')]
+    fwd_in = []
     fwd_buffer = []
     fwd_out = [torch.zeros_like(output_op_data.data, device='meta')]
 

colossalai/auto_parallel/meta_profiler/metainfo.py

@@ -1,6 +1,5 @@
 from typing import Callable, List
 
-import numpy as np
 import torch
 
 from colossalai.auto_parallel.tensor_shard.sharding_strategy import (
@@ -13,7 +12,7 @@ from colossalai.auto_parallel.tensor_shard.sharding_strategy import (
 )
 from colossalai.tensor.sharding_spec import ShardingSpec
 
-from .constants import INPLACE_MODULE, NO_SAVE_ACTIVATION
+from .constants import INPLACE_MODULE, INPLACE_OPS, NO_SAVE_ACTIVATION
 from .registry import meta_register
 
 __all__ = ['MetaInfo']
@@ -71,25 +70,12 @@ class MetaInfo:
         if self._strategy is not None and self._target is not None:
             self.compute_metainfo()
 
-    def compute_sharded_tensor(self, operation_data: OperationData, sharding_spec: ShardingSpec) -> torch.Tensor:
+    def compute_sharded_opdata(self, operation_data: OperationData, sharding_spec: ShardingSpec) -> torch.Tensor:
         """
-        Compute sharded meta tensor based on the given data and sharding spec.
+        Compute sharded opdata based on the given data and sharding spec.
         """
-        shard_sequnce = sharding_spec.sharding_sequence
-        device_mesh = sharding_spec.device_mesh
-        shape = operation_data.data.shape
-
-        new_shape = []
-        for dim, shard in zip(shape, shard_sequnce):
-            if shard.is_replica:
-                # replica
-                new_shape.append(dim)
-            else:
-                # sharded according to device_mesh shape
-                new_shape.append(dim // np.prod(np.array([device_mesh.mesh_shape[i] for i in shard.shard_list])))
-
         return OperationData(name=operation_data.name,
-                             data=torch.zeros(new_shape, device="meta"),
+                             data=torch.zeros(sharding_spec.get_sharded_shape_per_device(), device="meta"),
                              type=operation_data.type,
                              logical_shape=operation_data.logical_shape)
 
@@ -113,11 +99,13 @@ class MetaInfo:
             save_fwd_in = self._target.__class__ not in NO_SAVE_ACTIVATION
 
         # construct args for meta_func
-        args = [self.compute_sharded_tensor(k, v) for k, v in self._strategy.sharding_specs.items()]
+        args = [self.compute_sharded_opdata(k, v) for k, v in self._strategy.sharding_specs.items()]
 
         # construct kwargs
         if self.target in INPLACE_MODULE:
             kwargs = {'inplace': self.target.inplace}
+        elif self.target in INPLACE_OPS:
+            kwargs = {'inplace': True}
         else:
             kwargs = {'inplace': False}
 

colossalai/auto_parallel/passes/comm_metainfo_pass.py

@@ -0,0 +1,113 @@
+from typing import Dict
+
+import torch
+from torch.fx import GraphModule
+from torch.fx.node import Node
+
+from colossalai.auto_parallel.meta_profiler import MetaInfo
+from colossalai.auto_parallel.passes.runtime_apply_pass import runtime_apply, runtime_comm_spec_apply
+from colossalai.auto_parallel.tensor_shard.sharding_strategy import MemoryCost, TrainCycleItem
+from colossalai.tensor.comm_spec import CommSpec
+from colossalai.tensor.shape_consistency import ShapeConsistencyManager
+from colossalai.tensor.sharding_spec import ShardingSpec
+
+shape_consistency_manager = ShapeConsistencyManager()
+
+
+def _construct_meta_info(node: Node, origin_sharding_spec: ShardingSpec,
+                         target_sharding_spec: ShardingSpec) -> MetaInfo:
+    # get comm_action_sequence and total_cost from shape_consistency_manager
+    _, comm_action_sequence, total_cost = shape_consistency_manager.shape_consistency(
+        origin_sharding_spec, target_sharding_spec)
+
+    meta_info = MetaInfo()
+    # NOTE: the cost in shape_consistency_manager.mem_cost is the count in number of numel
+    # get mem cost for MetaInfo
+    mem_cost = shape_consistency_manager.mem_cost(comm_action_sequence)
+    # extract user that has _meta_data and extract element length
+    input_node = next(n for n in node._input_nodes if hasattr(n, '_meta_data'))
+    element_length = input_node._meta_data.element_size()
+
+    mem_cost.fwd.activation *= element_length
+    mem_cost.fwd.temp *= element_length
+    mem_cost.bwd.activation *= element_length
+    mem_cost.bwd.temp *= element_length
+    mem_cost.total.activation *= element_length
+
+    meta_info.memory_cost = mem_cost
+
+    # get computation cost for MetaInfo
+    meta_info.compute_cost = TrainCycleItem(total_cost['forward'] * element_length,
+                                            total_cost['backward'] * element_length,
+                                            total_cost['total'] * element_length)
+
+    # get tensor shape for MetaInfo
+    origin_sharding_spec: ShardingSpec
+    target_sharding_spec: ShardingSpec
+    input_shape = origin_sharding_spec.get_sharded_shape_per_device()
+    output_shape = target_sharding_spec.get_sharded_shape_per_device()
+
+    meta_info.fwd_in = [torch.rand(input_shape, device='meta')]
+    meta_info.fwd_buffer = []
+    meta_info.fwd_out = [torch.rand(output_shape, device='meta')]
+
+    return meta_info
+
+
+def _runtime_apply_meta_info(node: Node, origin_spec_dict, sharding_spec_dict) -> MetaInfo:
+    """
+    This method is used to construct `MetaInto` for shape consistency node
+    """
+
+    # extract node index and user node index
+    args = node.args
+    node_index, user_node_index = args[3], args[4]
+    origin_sharding_spec, target_sharding_spec = origin_spec_dict[node_index], sharding_spec_dict[node_index][
+        user_node_index]
+
+    return _construct_meta_info(node, origin_sharding_spec, target_sharding_spec)
+
+
+def _runtime_comm_spec_apply_meta_info(node: Node, comm_actions_dict: Dict) -> MetaInfo:
+    # extract node_index and op_data_name
+    node_index, op_data_name = node.args[2], node.args[3]
+
+    comm_action = comm_actions_dict[node_index][op_data_name]
+    if isinstance(comm_action.comm_spec, CommSpec):
+        # this case is for all_reduce, there will be no memory cost
+        meta_info = MetaInfo()
+        meta_info.memory_cost = TrainCycleItem(MemoryCost(), MemoryCost(), MemoryCost)
+        output_node = next(n for n in node.users if hasattr(n, '_meta_data'))
+        element_length = output_node._meta_data.element_size()
+
+        total_cost = comm_action.comm_spec.get_comm_cost()
+        meta_info.compute_cost = TrainCycleItem(total_cost['forward'] * element_length,
+                                                total_cost['backward'] * element_length,
+                                                total_cost['total'] * element_length)
+
+        input_shape = output_shape = comm_action.comm_spec.sharding_spec.get_sharded_shape_per_device()
+        meta_info.fwd_in = [torch.rand(input_shape, device='meta')]
+        meta_info.fwd_buffer = []
+        meta_info.fwd_out = [torch.rand(output_shape, device='meta')]
+    else:
+        # this case will be handled by shape consistency manager
+        origin_sharding_spec, target_sharding_spec = comm_action.comm_spec['src_spec'], comm_action.comm_spec[
+            'tgt_spec']
+        meta_info = _construct_meta_info(node, origin_sharding_spec, target_sharding_spec)
+
+    return meta_info
+
+
+def comm_metainfo_pass(gm: GraphModule, sharding_spec_dict: Dict, origin_spec_dict: Dict,
+                       comm_actions_dict: Dict) -> GraphModule:
+    """
+    The method manages all the metainfo of the communication node (run_time_apply, runtime_comm_spec_apply) in the graph.
+    """
+    for node in gm.graph.nodes:
+        if node.target == runtime_apply:
+            setattr(node, 'best_metainfo', _runtime_apply_meta_info(node, origin_spec_dict, sharding_spec_dict))
+        elif node.target == runtime_comm_spec_apply:
+            setattr(node, 'best_metainfo', _runtime_comm_spec_apply_meta_info(node, comm_actions_dict))
+        else:
+            pass
+    return gm

colossalai/auto_parallel/passes/constants.py

@@ -0,0 +1,8 @@
+import torch
+
+OUTPUT_SAVED_OPS = [torch.nn.functional.relu, torch.nn.functional.softmax, torch.flatten]
+
+OUTPUT_SAVED_MOD = [
+    torch.nn.ReLU,
+    torch.nn.Softmax,
+]

colossalai/auto_parallel/passes/meta_info_prop.py

@@ -1,17 +1,16 @@
 import uuid
 from dataclasses import asdict
-from typing import Any, Dict, List, NamedTuple, Tuple
+from typing import List
 
 import torch
 import torch.fx
 from torch.fx import GraphModule
-from torch.fx.node import Argument, Node, Target
+from torch.fx.node import Node
-from torch.utils._pytree import tree_map
 
 from colossalai.auto_parallel.meta_profiler import MetaInfo
-from colossalai.fx._compatibility import compatibility, is_compatible_with_meta
+from colossalai.auto_parallel.passes.constants import OUTPUT_SAVED_MOD, OUTPUT_SAVED_OPS
+from colossalai.fx._compatibility import compatibility
 from colossalai.fx.profiler import GraphInfo
-from colossalai.fx.profiler.constants import OUTPUT_SAVED_MOD, OUTPUT_SAVED_OPS
 
 
 def _normalize_tuple(x):
@@ -47,7 +46,7 @@ class MetaInfoProp:
         """
         Check if the node is inplace operation.
         """
-        if node.op == 'call_method':
+        if node.op == 'call_module':
             return node.graph.owning_module.get_submodule(node.target).__class__ in OUTPUT_SAVED_MOD
         elif node.op == "call_function":
             return node.target in OUTPUT_SAVED_OPS
@@ -68,7 +67,7 @@ class MetaInfoProp:
         """
         graph_info = GraphInfo()
         out = _normalize_tuple(getattr(node, '_meta_data', None))
-        graph_info.fwd_out = list(out)
+        graph_info.fwd_out = list(out) if out[0] is not None else []
         node.meta = {**asdict(graph_info)}
 
     @compatibility(is_backward_compatible=False)
@@ -97,66 +96,70 @@ class MetaInfoProp:
         """
         Handle other kind of nodes
         """
-        assert hasattr(node, 'best_metainfo'), f"Cannot find best_metainfo in node {node}"
+        assert hasattr(node, 'best_metainfo'), f"Cannot find best_metainfo in node {node}, {node.op}"
         graph_info = GraphInfo()
         meta_info = node.best_metainfo
         meta_info: MetaInfo
 
         # set data_ptr for input_tensor in MetaInfo class
-        input_tensor: List[torch.Tensor] = meta_info.fwd_in
+        input_tensors: List[torch.Tensor] = meta_info.fwd_in
-        buffer_tensor: List[torch.Tensor] = meta_info.fwd_buffer
+        buffer_tensors: List[torch.Tensor] = meta_info.fwd_buffer
-        output_tensor: List[torch.Tensor] = meta_info.fwd_out
+        output_tensors: List[torch.Tensor] = meta_info.fwd_out
 
-        if len(input_tensor) > 0:
+        if self._is_inplace(node):
+            # inplace operation will not create new tensor, and it only has one parent node
+            # TODO: Verify this observation
+            # set data_ptr for input_tensor, buffer_tensor and output_tensor of current node
+            parent_node = list(node._input_nodes.keys())[0]
+            parent_tensor = parent_node.meta.get("fwd_out")[0]
+            parent_tensor: torch.Tensor
+            for tensor in input_tensors:
+                tensor.data_ptr = parent_tensor.data_ptr
+            for tensor in buffer_tensors:
+                tensor.data_ptr = parent_tensor.data_ptr
+            for tensor in output_tensors:
+                tensor.data_ptr = parent_tensor.data_ptr
+
+        else:
             for par in node._input_nodes:
-                if par.meta:
+                # set data_ptr for the input_tensor of current node from the output_tensor of its parent node
-                    if len(par.meta["fwd_out"]) > 0:
+                for tensor in par.meta.get("fwd_out", []):
-                        # set data_ptr for the input_tensor of current node from the output_tensor of its parent node
+                    tensor: torch.Tensor
-                        for tensor in par.meta["fwd_out"]:
+                    target_input_tensor = next(
-                            tensor: torch.Tensor
+                        (x for x in input_tensors if not x.data_ptr() and x.shape == tensor.shape), None)
-                            target_tensor = next(
+                    if target_input_tensor is not None:
-                                (x for x in input_tensor if not x.data_ptr() and x.shape == tensor.shape), None)
+                        target_input_tensor.data_ptr = tensor.data_ptr
-                            target_tensor.data_ptr = tensor.data_ptr
 
             # set data_ptr for tensor in input_tensor that is not set
-            for tensor in input_tensor:
+            for tensor in input_tensors:
                 if not tensor.data_ptr():
                     self._set_data_ptr(tensor)
 
-        # attach it to graph_info
-        graph_info.fwd_in = input_tensor
-
-        if self._is_inplace(node):
-            # inplace operation will not create new tensor
-            # set data_ptr for buffer_tensor and output_tensor of current node
-            for tensor in input_tensor:
-                tensor: torch.Tensor
-                target_buffer_tensor = next((x for x in buffer_tensor if not x.data_ptr() and x.shape == tensor.shape),
-                                            None)
-                target_output_tensor = next((x for x in output_tensor if not x.data_ptr() and x.shape == tensor.shape),
-                                            None)
-                target_buffer_tensor.data_ptr = tensor.data_ptr
-                target_output_tensor.data_ptr = tensor.data_ptr
-            # attach them to graph_info
-            graph_info.fwd_tmp = buffer_tensor
-            graph_info.fwd_out = output_tensor
-
-        else:
             # set data_ptr for buffer_tensor
-            for tensor in buffer_tensor:
+            for tensor in buffer_tensors:
                 self._set_data_ptr(tensor)
-            # attach it to graph_info
-            graph_info.fwd_tmp = buffer_tensor
 
             # set data_ptr for output_tensor
-            for tensor in output_tensor:
+            for tensor in output_tensors:
                 self._set_data_ptr(tensor)
-            # attach it to graph_info
+
-            graph_info.fwd_out = output_tensor
+        # attach them to graph_info
+        graph_info.fwd_in = input_tensors
+        graph_info.fwd_tmp = buffer_tensors
+        graph_info.fwd_out = output_tensors
 
         # fetch other memory informations
         memory_cost = meta_info.memory_cost
         graph_info.fwd_mem_tmp = memory_cost.fwd.temp
+        graph_info.fwd_mem_out = memory_cost.fwd.activation
         graph_info.bwd_mem_tmp = memory_cost.bwd.temp
+        graph_info.bwd_mem_out = memory_cost.bwd.activation
+
+        # fetch flop information
+        # here we use fwd_time and bwd_time to deal with the case that
+        # communication cost is a float
+        compute_cost = meta_info.compute_cost
+        graph_info.fwd_time = compute_cost.fwd
+        graph_info.bwd_time = compute_cost.bwd
 
         node.meta = {**asdict(graph_info)}

colossalai/auto_parallel/passes/runtime_apply_pass.py

@@ -47,53 +47,6 @@ def runtime_apply_for_iterable_object(node: Node, origin_dict: Dict, input_dict:
     return rst
 
 
-def construct_meta_info(node: Node, user_node: Node) -> MetaInfo:
-    """
-    This method is used to construct `MetaInto` for shape consistency node
-    TODO: Actually we could attain the cost information from resharding cost in node
-    handler, we should modify this part in the future.
-    """
-
-    def compute_shape(sharding_spec: ShardingSpec):
-        shape = sharding_spec.entire_shape
-        new_shape = []
-        for dim, shard in sharding_spec.dim_partition_dict.items():
-            new_shape.append(shape[dim] // len(shard))
-        return new_shape
-
-    meta_info = MetaInfo()
-    origin_sharding_spec, target_sharding_spec = node.sharding_spec, user_node.best_strategy.get_sharding_spec_by_name(
-        str(node.name))
-    _, comm_action_sequence, total_cost = shape_consistency_manager.shape_consistency(
-        origin_sharding_spec, target_sharding_spec)
-
-    # NOTE: the cost in shape_consistency_manager.mem_cost is the count in number of numel
-    # get mem cost for MetaInfo
-    mem_cost = shape_consistency_manager.mem_cost(comm_action_sequence)
-    element_length = node._meta_data.element_size()
-    mem_cost.fwd.activation *= element_length
-    mem_cost.fwd.temp *= element_length
-    mem_cost.bwd.activation *= element_length
-    mem_cost.bwd.temp *= element_length
-    mem_cost.total.activation *= element_length
-
-    meta_info.memory_cost = mem_cost
-
-    # get computation cost for MetaInfo
-    compute_cost = TrainCycleItem(total_cost['forward'], total_cost['backward'], total_cost['total'])
-    meta_info.compute_cost = compute_cost
-
-    # get tensor shape for MetaInfo
-    input_shape = compute_shape(origin_sharding_spec)
-    output_shape = compute_shape(target_sharding_spec)
-
-    meta_info.fwd_in = [torch.rand(input_shape, device='meta')]
-    meta_info.fwd_buffer = []
-    meta_info.fwd_out = [torch.rand(output_shape, device='meta')]
-
-    return meta_info
-
-
 def runtime_comm_spec_apply(tensor: torch.Tensor, comm_actions_dict: Dict, node_index: int, op_data_name: str):
     """
     This method will be invoked during runtime to apply the comm action following the instruction of comm spec.
@@ -175,8 +128,6 @@ def _shape_consistency_apply(gm: torch.fx.GraphModule):
                                                                    runtime_apply,
                                                                    args=(node, origin_dict_node, input_dict_node,
                                                                          node_to_index_dict[node], user_node_index))
-                    meta_info = construct_meta_info(node, user_node)
-                    setattr(shape_consistency_node, 'best_metainfo', meta_info)
 
             new_args = list(user_node.args)
             new_kwargs = dict(user_node.kwargs)

colossalai/auto_parallel/tensor_shard/node_handler/binary_elementwise_handler.py

@@ -16,7 +16,7 @@ __all__ = ['BinaryElementwiseHandler']
 
 
 @operator_registry.register(BCAST_FUNC_OP)
-class BinaryElementwiseHandler(NodeHandler):
+class BinaryElementwiseHandler(MetaInfoNodeHandler):
     """
     An BinaryBcastOpHandler is a node handler which deals with operations which have two
     operands and broadcasting occurs such as torch.add.

colossalai/auto_parallel/tensor_shard/node_handler/node_handler.py

@@ -4,7 +4,7 @@ from typing import Dict, List, Tuple, Union
 import torch
 from torch.fx.node import Node
 
-from colossalai.auto_parallel.meta_profiler.metainfo import MetaInfo
+from colossalai.auto_parallel.meta_profiler.metainfo import MetaInfo, meta_register
 from colossalai.auto_parallel.tensor_shard.sharding_strategy import (
     OperationData,
     OperationDataType,
@@ -138,8 +138,7 @@ class NodeHandler(ABC):
             return None
 
         if self.node.op == 'call_module':
-            submod = self.node.graph.owning_module.get_submodule(self.node.target)
+            target = self.node.graph.owning_module.get_submodule(self.node.target)
-            target = type(submod)
         elif self.node.op == 'call_function':
             target = self.node.target
         elif self.node.op == 'call_method':
@@ -235,15 +234,19 @@ class MetaInfoNodeHandler(NodeHandler):
         """
         super().register_strategy(compute_resharding_cost=compute_resharding_cost)
         target = self.get_target_function()
-        metainfo_vector = []
+        # Currently we haven't patched all the torch functions and modules, so if the target
-        for strategy in self.strategies_vector:
+        # is not patched, we will use the default cost model to compute the cost.
-            metainfo = MetaInfo(strategy, target)
+        # TODO: patch all torch functions and modules to make it clean
-            strategy.compute_cost = metainfo.compute_cost
+        if meta_register.has(target.__class__) or meta_register.has(target):
-            strategy.memory_cost = metainfo.memory_cost
+            metainfo_vector = []
-            metainfo_vector.append(metainfo)
+            for strategy in self.strategies_vector:
+                metainfo = MetaInfo(strategy, target)
+                strategy.compute_cost = metainfo.compute_cost
+                strategy.memory_cost = metainfo.memory_cost
+                metainfo_vector.append(metainfo)
 
-        # attach metainfos to the handler
+            # attach metainfos to the handler
-        setattr(self, "metainfo_vector", metainfo_vector)
+            setattr(self, "metainfo_vector", metainfo_vector)
 
         return self.strategies_vector
 
@@ -282,14 +285,18 @@ class MetaInfoModuleHandler(ModuleHandler):
         """
         super().register_strategy(compute_resharding_cost=compute_resharding_cost)
         target = self.get_target_function()
-        metainfo_vector = []
+        # Currently we haven't patched all the torch functions and modules, so if the target
-        for strategy in self.strategies_vector:
+        # is not patched, we will use the default cost model to compute the cost.
-            metainfo = MetaInfo(strategy, target)
+        # TODO: patch all torch functions and modules to make it clean
-            strategy.compute_cost = metainfo.compute_cost
+        if meta_register.has(target.__class__) or meta_register.has(target):
-            strategy.memory_cost = metainfo.memory_cost
+            metainfo_vector = []
-            metainfo_vector.append(metainfo)
+            for strategy in self.strategies_vector:
+                metainfo = MetaInfo(strategy, target)
+                strategy.compute_cost = metainfo.compute_cost
+                strategy.memory_cost = metainfo.memory_cost
+                metainfo_vector.append(metainfo)
 
-        # attach metainfos to the handler
+            # attach metainfos to the handler
-        setattr(self, "metainfo_vector", metainfo_vector)
+            setattr(self, "metainfo_vector", metainfo_vector)
 
         return self.strategies_vector

colossalai/auto_parallel/tensor_shard/node_handler/reshape_handler.py

@@ -3,7 +3,7 @@ from typing import Dict, List
 import torch
 
 from ..sharding_strategy import OperationData, OperationDataType
-from .node_handler import NodeHandler
+from .node_handler import MetaInfoNodeHandler, NodeHandler
 from .registry import operator_registry
 from .strategy import ReshapeGenerator, StrategyGenerator
 
@@ -13,7 +13,7 @@ __all__ = ['ReshapeHandler']
 @operator_registry.register(torch.flatten)
 @operator_registry.register(torch.Tensor.unsqueeze)
 @operator_registry.register(torch.nn.AdaptiveAvgPool2d)
-class ReshapeHandler(NodeHandler):
+class ReshapeHandler(MetaInfoNodeHandler):
     """
     A ReshapeHandler which deals with the sharding strategies for Reshape Op, such as torch.reshape.
     """

colossalai/auto_parallel/tensor_shard/node_handler/unary_elementwise_handler.py

@@ -3,7 +3,7 @@ from typing import Dict, List
 import torch
 
 from ..sharding_strategy import OperationData, OperationDataType
-from .node_handler import NodeHandler
+from .node_handler import MetaInfoNodeHandler, NodeHandler
 from .registry import operator_registry
 from .strategy import StrategyGenerator, UnaryElementwiseGenerator
 
@@ -19,7 +19,7 @@ __all__ = ['UnaryElementwiseHandler']
 @operator_registry.register(torch.nn.modules.dropout.Dropout)
 @operator_registry.register(torch.Tensor.contiguous)
 @operator_registry.register(torch.nn.functional.dropout)
-class UnaryElementwiseHandler(NodeHandler):
+class UnaryElementwiseHandler(MetaInfoNodeHandler):
     """
     A UnaryElementwiseHandler which deals with the sharding strategies for UnaryElementwise Op.
     """

colossalai/fx/profiler/shard_utils.py

@@ -100,7 +100,7 @@ def calculate_fwd_time(n: Node) -> float:
         fwd_time (float): the result of `fwd_time`
     """
     # TODO(super-dainiu): should divide the time by the number of GPUs as well as TFLOPs
-    return n.meta["fwd_flop"]
+    return n.meta["fwd_time"]
 
 
 def calculate_bwd_time(n: Node) -> float:
@@ -111,4 +111,4 @@ def calculate_bwd_time(n: Node) -> float:
         bwd_time (float): the result of `bwd_time`
     """
     # TODO(super-dainiu): should divide the time by the number of GPUs as well as TFLOPs
-    return n.meta["bwd_flop"]
+    return n.meta["bwd_time"]

colossalai/tensor/shape_consistency.py

@@ -441,6 +441,8 @@ class ShapeConsistencyManager(metaclass=SingletonMeta):
             if discard_input:
                 alloc_numel -= input_numel
 
+            return alloc_numel, peak_numel
+
         def split_analysis(comm_spec: CommSpec, discard_input: bool, alloc_numel: int, peak_numel: int):
             """analyze split memory footprint
             split will allocate memory for the output tensor if we don't apply shard on the first dimension of
@@ -478,11 +480,13 @@ class ShapeConsistencyManager(metaclass=SingletonMeta):
                 # kind of weird, and I think we could ignore it for now.
                 pass
 
+            return alloc_numel, peak_numel
+
         def reduce_analysis(comm_spec: CommSpec, discard_input: bool, alloc_numel: int, peak_numel: int):
             """
             a dummy function for reduce memory footprint analysis, as the reduce action doesn't allocate extra memory
             """
-            pass
+            return alloc_numel, peak_numel
 
         def all2all_analysis(comm_spec: CommSpec, discard_input: bool, alloc_numel: int, peak_numel: int):
             """analyze all_to_all memory footprint
@@ -508,11 +512,13 @@ class ShapeConsistencyManager(metaclass=SingletonMeta):
             if discard_input:
                 alloc_numel -= input_numel
 
+            return alloc_numel, peak_numel
+
         def identity_analysis(comm_spec: CommSpec, discard_input: bool, alloc_numel: int, peak_numel: int):
             """
             a dummy function for identity memory footprint analysis, as the identity action doesn't allocate extra memory
             """
-            pass
+            return alloc_numel, peak_numel
 
         pattern_to_func_dict = {
             CollectiveCommPattern.GATHER_FWD_SPLIT_BWD: [gather_analysis, split_analysis],
@@ -539,17 +545,18 @@ class ShapeConsistencyManager(metaclass=SingletonMeta):
         for idx, action_spec_pair in enumerate(zip(fwd_actions, comm_action_sequence)):
             # the first forward comm action will not discard input
             fwd_action, comm_spec = action_spec_pair
-            if idx == 0:
+            fwd_alloc_numel, fwd_peak_numel = fwd_action(comm_spec, False, fwd_alloc_numel,
-                fwd_action(comm_spec, False, fwd_alloc_numel, fwd_peak_numel)
+                                                         fwd_peak_numel) if idx == 0 else fwd_action(
-            else:
+                                                             comm_spec, True, fwd_alloc_numel, fwd_peak_numel)
-                fwd_action(comm_spec, True, fwd_alloc_numel, fwd_peak_numel)
 
         # analyze memory footprint for backward comm actions sequence
         bwd_alloc_numel = 0
         bwd_peak_numel = 0
         for idx, action_spec_pair in enumerate(zip(reversed(bwd_actions), reversed(comm_action_sequence))):
             bwd_action, comm_spec = action_spec_pair
-            bwd_action(comm_spec, True, bwd_alloc_numel, bwd_peak_numel)
+            bwd_alloc_numel, bwd_peak_numel = bwd_action(comm_spec, False, bwd_alloc_numel,
+                                                         bwd_peak_numel) if idx == 0 else bwd_action(
+                                                             comm_spec, True, bwd_alloc_numel, bwd_peak_numel)
 
         fwd_mem = MemoryCost(activation=fwd_alloc_numel, temp=fwd_peak_numel - fwd_alloc_numel)
         bwd_mem = MemoryCost(activation=bwd_alloc_numel, temp=bwd_peak_numel - bwd_alloc_numel)