[autoparallel] adapt handlers with attention block (#1990)

* [autoparallel] adapt handlers with attention block

* polish

colossalai/auto_parallel/tensor_shard/node_handler/reshape_handler.py

@@ -12,6 +12,7 @@ __all__ = ['ReshapeHandler']
 
 @operator_registry.register(torch.reshape)
 @operator_registry.register(torch.Tensor.split)
+@operator_registry.register(torch.split)
 @operator_registry.register(torch.flatten)
 @operator_registry.register(torch.Tensor.transpose)
 @operator_registry.register(torch.Tensor.permute)

colossalai/auto_parallel/tensor_shard/node_handler/strategy/batch_norm_generator.py

@@ -220,7 +220,9 @@ class BatchNormStrategyGenerator(StrategyGenerator):
             logical_process_axis=mesh_dim_0,
             comm_type=CommType.IMPLICIT)
 
-        communication_action_mapping = {"output": output_comm_action}
+        # TODO: Temporary solution has no communication cost,
+        # above action should be added after the SyncBN replace pass completed.
+        communication_action_mapping = {}
 
         return self.get_sharding_strategy(name=name,
                                           sharding_spec_mapping=sharding_spec_mapping,
@@ -256,7 +258,9 @@ class BatchNormStrategyGenerator(StrategyGenerator):
             logical_process_axis=[mesh_dim_0, mesh_dim_1],
             comm_type=CommType.IMPLICIT)
 
-        communication_action_mapping = {"output": output_comm_action}
+        # TODO: Temporary solution has no communication cost,
+        # above action should be added after the SyncBN replace pass completed.
+        communication_action_mapping = {}
 
         return self.get_sharding_strategy(name=name,
                                           sharding_spec_mapping=sharding_spec_mapping,
@@ -302,7 +306,9 @@ class BatchNormStrategyGenerator(StrategyGenerator):
             logical_process_axis=[mesh_dim_0],
             comm_type=CommType.IMPLICIT)
 
-        communication_action_mapping = {"output": output_comm_action}
+        # TODO: Temporary solution has no communication cost,
+        # above action should be added after the SyncBN replace pass completed.
+        communication_action_mapping = {}
 
         return self.get_sharding_strategy(name=name,
                                           sharding_spec_mapping=sharding_spec_mapping,

colossalai/auto_parallel/tensor_shard/node_handler/strategy/getitem_generator.py

@@ -69,7 +69,7 @@ class TensorStrategyGenerator(GetItemStrategyGenerator):
 
     def collate_strategies(self) -> List[ShardingStrategy]:
         strategy_list = []
-        for strategy in self.predecessor_node.strategies_vector:
+        for index, strategy in enumerate(self.predecessor_node.strategies_vector):
             dim_partition_dict_mapping = {}
             communication_action_mapping = {}
             dim_partition_dict_for_input = strategy.output_sharding_specs[self.op_data["input"]].dim_partition_dict
@@ -96,7 +96,7 @@ class TensorStrategyGenerator(GetItemStrategyGenerator):
                     arg_index=0)
                 communication_action_mapping["input"] = input_communication_action
 
-            name = f'{sharding_spec_mapping["output"].sharding_sequence} = {sharding_spec_mapping["input"].sharding_sequence}'
+            name = f'{sharding_spec_mapping["output"].sharding_sequence} = {sharding_spec_mapping["input"].sharding_sequence}_{index}'
 
             strategy = self.get_sharding_strategy(name=name,
                                                   sharding_spec_mapping=sharding_spec_mapping,
@@ -121,7 +121,7 @@ class TensorTupleStrategyGenerator(GetItemStrategyGenerator):
         strategy_list = []
         index = self.op_data["index"].data
 
-        for strategy in self.predecessor_node.strategies_vector:
+        for strategy_index, strategy in enumerate(self.predecessor_node.strategies_vector):
             # the sharding spec for input in this case is a tuple of ShardingSpec.
             sharding_spec_for_input = strategy.output_sharding_specs[self.op_data["input"]]
             dim_partition_dict_for_output = sharding_spec_for_input[index].dim_partition_dict
@@ -132,8 +132,11 @@ class TensorTupleStrategyGenerator(GetItemStrategyGenerator):
             }
             sharding_spec_mapping = self.to_sharding_spec_mapping(dim_partition_dict_mapping)
             sharding_spec_mapping["input"] = sharding_spec_for_input
-
+            input_sharding_info = f"get the {index} element from ("
-            name = f'{sharding_spec_mapping["output"].sharding_sequence} = {sharding_spec_mapping["input"].sharding_sequence}'
+            for sharding_spec in sharding_spec_for_input:
+                input_sharding_info += f'{sharding_spec.sharding_sequence}, '
+            input_sharding_info += ")"
+            name = f'{sharding_spec_mapping["output"].sharding_sequence} = {input_sharding_info}_{strategy_index}'
 
             strategy = self.get_sharding_strategy(name=name,
                                                   sharding_spec_mapping=sharding_spec_mapping,

colossalai/auto_parallel/tensor_shard/node_handler/strategy/where_generator.py

@@ -1,9 +1,12 @@
 import copy
 from typing import List
 
-from colossalai.auto_parallel.tensor_shard.sharding_strategy import (MemoryCost, ShardingStrategy, TrainCycleItem)
+from colossalai.auto_parallel.tensor_shard.sharding_strategy import MemoryCost, ShardingStrategy, TrainCycleItem
-from colossalai.auto_parallel.tensor_shard.utils import (enumerate_all_possible_1d_sharding,
+from colossalai.auto_parallel.tensor_shard.utils import (
-                                                         enumerate_all_possible_2d_sharding)
+    enumerate_all_possible_1d_sharding,
+    enumerate_all_possible_2d_sharding,
+    ignore_sharding_exception,
+)
 
 from .strategy_generator import StrategyGenerator
 
@@ -50,6 +53,7 @@ class WhereGenerator(StrategyGenerator):
         memory_cost = TrainCycleItem(fwd=fwd_mem_cost, bwd=bwd_mem_cost, total=total_mem_cost)
         strategy.memory_cost = memory_cost
 
+    @ignore_sharding_exception
     def _generate_strategy_with_dim_partition(self, dim_partition):
         dim_partition_dict_mapping = {
             "condition": dim_partition,

colossalai/auto_parallel/tensor_shard/node_handler/unary_elementwise_handler.py

@@ -14,6 +14,11 @@ __all__ = ['UnaryElementwiseHandler']
 @operator_registry.register(torch.Tensor.type)
 @operator_registry.register(torch.abs)
 @operator_registry.register(torch.nn.ReLU)
+# TODO: softmax need to be relocated
+@operator_registry.register(torch.nn.functional.softmax)
+@operator_registry.register(torch.nn.modules.dropout.Dropout)
+@operator_registry.register(torch.Tensor.contiguous)
+@operator_registry.register(torch.nn.functional.dropout)
 class UnaryElementwiseHandler(NodeHandler):
     """
     A UnaryElementwiseHandler which deals with the sharding strategies for UnaryElementwise Op.

colossalai/auto_parallel/tensor_shard/node_handler/where_handler.py

@@ -57,24 +57,6 @@ class WhereHandler(NodeHandler):
         logical_operand.logical_shape = target_shape
         return logical_operand
 
-    def register_strategy(self, compute_resharding_cost: bool = False) -> StrategiesVector:
-        """
-        Register different sharding strategies for the current node.
-        """
-        strategy_generators = self.get_strategy_generator()
-
-        for generator in strategy_generators:
-            strategies = generator.generate()
-            strategies_vector = map(self.post_process, strategies)
-            # compute the resharding costs based on the previous node
-            # strategies if specified
-            if compute_resharding_cost:
-                strategies = list(map(self.update_resharding_cost, strategies))
-            self.strategies_vector.extend(strategies)
-
-        self.strategies_vector = list(strategies_vector)
-        return self.strategies_vector
-
     def post_process(self, strategy: ShardingStrategy):
         logical_op_data_mapping, physical_op_data_mapping = self.get_operation_data_mapping()
         for key in logical_op_data_mapping.keys():

tests/test_auto_parallel/test_tensor_shard/test_node_handler/test_getitem_handler.py

@@ -3,6 +3,8 @@ import torch.nn as nn
 
 from colossalai.auto_parallel.tensor_shard.node_handler.conv_handler import ConvFunctionHandler
 from colossalai.auto_parallel.tensor_shard.node_handler.getitem_handler import GetItemHandler
+from colossalai.auto_parallel.tensor_shard.node_handler.placeholder_handler import PlacehodlerHandler
+from colossalai.auto_parallel.tensor_shard.node_handler.reshape_handler import ReshapeHandler
 from colossalai.auto_parallel.tensor_shard.sharding_strategy import OperationData, OperationDataType, StrategiesVector
 from colossalai.device.device_mesh import DeviceMesh
 from colossalai.fx import ColoGraphModule, ColoTracer
@@ -10,7 +12,7 @@ from colossalai.fx.tracer.meta_patch.patched_module import linear
 from colossalai.testing.pytest_wrapper import run_on_environment_flag
 
 
-class GetItemModel(nn.Module):
+class GetItemFromTensorModel(nn.Module):
 
     def __init__(self):
         super().__init__()
@@ -21,8 +23,8 @@ class GetItemModel(nn.Module):
         return x
 
 
-def test_getitem_function_handler():
+def test_getitem_from_tensor_handler():
-    model = GetItemModel()
+    model = GetItemFromTensorModel()
     tracer = ColoTracer()
     # graph():
     #     %input_1 : torch.Tensor [#users=1] = placeholder[target=input]
@@ -83,5 +85,83 @@ def test_getitem_function_handler():
     assert len(getitem_strategies_vector) == len(conv_strategies_vector)
 
 
+class GetItemFromTupleModel(nn.Module):
+
+    def __init__(self):
+        super().__init__()
+
+    def forward(self, input):
+        split_node = torch.split(input, 2, 0)
+        x = split_node[1]
+        return x
+
+
+def test_getitem_from_tuple_handler():
+    model = GetItemFromTupleModel()
+    tracer = ColoTracer()
+    # graph():
+    #     %input_1 : torch.Tensor [#users=1] = placeholder[target=input]
+    #     %split : [#users=1] = call_function[target=torch.functional.split](args = (%conv2d, 2), kwargs = {dim: 0})
+    #     %getitem : [#users=1] = call_function[target=operator.getitem](args = (%split, 1), kwargs = {})
+    #     return getitem
+    graph = tracer.trace(model, meta_args={
+        "input": torch.rand(4, 4, 64, 64).to('meta'),
+    })
+    gm = ColoGraphModule(model, graph)
+    physical_mesh_id = torch.arange(0, 4)
+
+    mesh_shape = (2, 2)
+    device_mesh = DeviceMesh(physical_mesh_id, mesh_shape)
+    input_node = list(graph.nodes)[0]
+    split_node = list(graph.nodes)[1]
+    getitem_node = list(graph.nodes)[2]
+    input_strategies_vector = StrategiesVector(input_node)
+    getitem_strategies_vector = StrategiesVector(getitem_node)
+    split_strategies_vector = StrategiesVector(split_node)
+
+    # build handler
+    input_handler = PlacehodlerHandler(
+        node=input_node,
+        device_mesh=device_mesh,
+        strategies_vector=input_strategies_vector,
+        placeholder_option='replicated',
+    )
+    input_handler.register_strategy(compute_resharding_cost=False)
+    setattr(input_node, 'strategies_vector', input_strategies_vector)
+    split_handler = ReshapeHandler(node=split_node, device_mesh=device_mesh, strategies_vector=split_strategies_vector)
+    split_handler.register_strategy(compute_resharding_cost=False)
+    setattr(split_node, 'strategies_vector', split_strategies_vector)
+    getitem_handler = GetItemHandler(node=getitem_node,
+                                     device_mesh=device_mesh,
+                                     strategies_vector=getitem_strategies_vector)
+    getitem_handler.register_strategy(compute_resharding_cost=False)
+    setattr(getitem_node, 'strategies_vector', getitem_strategies_vector)
+
+    # check operation data mapping
+    mapping = getitem_handler.get_operation_data_mapping()
+
+    for name, op_data in mapping.items():
+        op_data: OperationData
+        # make sure they have valid values
+        assert op_data.data is not None
+
+    assert mapping['input'].name == "split"
+    assert mapping['input'].type == OperationDataType.ARG
+    assert mapping['input'].logical_shape == (torch.Size([2, 4, 64, 64]), torch.Size([2, 4, 64, 64]))
+
+    assert mapping['index'].name == "index"
+    assert isinstance(mapping['index'].data, int)
+    assert mapping['index'].type == OperationDataType.ARG
+
+    assert mapping['output'].name == "getitem"
+    assert mapping['output'].data.is_meta
+    assert mapping['output'].data.shape == torch.Size([2, 4, 64, 64])
+    assert mapping['output'].type == OperationDataType.OUTPUT
+
+    # getitem is a following strategy handler, so the number of strategies is equal to the predecessor node.
+    assert len(getitem_strategies_vector) == len(split_strategies_vector)
+
+
 if __name__ == '__main__':
-    test_getitem_function_handler()
+    test_getitem_from_tensor_handler()
+    test_getitem_from_tuple_handler()