[MetaSchedule] Developer Ergonomics Enhancement II

python/tvm/meta_schedule/space_generator/__init__.py

@@ -19,7 +19,7 @@
 Meta Schedule design space generators that generates design
 space for generation of measure candidates.
 """
-from .space_generator import SpaceGenerator, PySpaceGenerator
-from .space_generator_union import SpaceGeneratorUnion
-from .schedule_fn import ScheduleFn
 from .post_order_apply import PostOrderApply
+from .schedule_fn import SCH_FN_TYPE, ScheduleFn
+from .space_generator import PySpaceGenerator, SpaceGenerator
+from .space_generator_union import SpaceGeneratorUnion

python/tvm/meta_schedule/space_generator/schedule_fn.py

@@ -30,18 +30,17 @@
 if TYPE_CHECKING:
     from ..tune_context import TuneContext
 
+SCH_FN_TYPE = Union[  # pylint: disable=invalid-name
+    Callable[[Schedule], None],  # No output
+    Callable[[Schedule], Schedule],  # Single output
+    Callable[[Schedule], List[Schedule]],  # Multiple outputs
+]
+
 
 @derived_object
 class ScheduleFn(PySpaceGenerator):
     """A design space generator with design spaces specified by a schedule function."""
 
-    # Multiple cases of schedule functions supported
-    SCH_FN_TYPE = Union[
-        Callable[[Schedule], None],  # No output
-        Callable[[Schedule], Schedule],  # Single output
-        Callable[[Schedule], List[Schedule]],  # Multiple outputs
-    ]
-
     def __init__(self, sch_fn: SCH_FN_TYPE):
         """Constructor.
 

python/tvm/meta_schedule/tune.py

@@ -88,7 +88,7 @@ class TuneConfig(NamedTuple):
     search_strategy_config: Optional[Dict[str, Any]] = None
     logger_config: Optional[Dict[str, Any]] = None
 
-    def create_strategy(self, **kwargs):
+    def create_strategy(self):
         """Create search strategy from configuration"""
         cls_tbl = {
             "evolutionary": EvolutionarySearch,
@@ -111,7 +111,6 @@ def create_strategy(self, **kwargs):
         return cls_tbl[self.strategy](
             num_trials_per_iter=self.num_trials_per_iter,
             max_trials_per_task=max_trials_per_task,
-            **kwargs,
             **config,
         )
 

python/tvm/meta_schedule/tune_context.py

@@ -17,7 +17,7 @@
 """Meta Schedule tuning context."""
 
 import logging
-from typing import TYPE_CHECKING, Dict, List, Optional
+from typing import TYPE_CHECKING, Dict, List, Optional, Union
 
 from tvm import IRModule
 from tvm._ffi import register_object
@@ -36,7 +36,8 @@
     from .runner import RunnerResult
     from .schedule_rule import ScheduleRule
     from .search_strategy import MeasureCandidate, SearchStrategy
-    from .space_generator import SpaceGenerator
+    from .space_generator import SCH_FN_TYPE, ScheduleFn, SpaceGenerator
+    from .tune import TuneConfig
 
 
 @register_object("meta_schedule.TuneContext")
@@ -54,16 +55,24 @@ class TuneContext(Object):
         The workload to be optimized.
     target : Optional[Target] = None
         The target to be optimized for.
-    space_generator : Optional[SpaceGenerator] = None
+    space_generator : Union[None, SCH_FN_TYPE, SpaceGenerator] = None
         The design space generator.
-    search_strategy : Optional[SearchStrategy] = None
+    search_strategy : Union[None, TuneConfig, SearchStrategy] = None
         The search strategy.
-    sch_rules: Optional[List[ScheduleRule]] = None,
+        if None, the strategy is left blank.
+        If TuneConfig, the strategy is initialized with the TuneConfig.create_strategy().
+    sch_rules: Union[None, str, List[ScheduleRule]] = None,
         The schedule rules.
-    postprocs: Optional[List[Postproc"]] = None,
+        If None, use an empty list of rules.
+        if "default", use target-default rules.
+    postprocs: Union[None, str, List[Postproc"]] = None,
         The postprocessors.
-    mutator_probs: Optional[Dict[Mutator, float]]
+        If None, use an empty list of rules.
+        if "default", use target-default rules.
+    mutator_probs: Union[None, str, Dict[Mutator, float]]
         Mutators and their probability mass.
+        If None, use an empty list of rules.
+        if "default", use target-default rules.
     task_name : Optional[str] = None
         The name of the tuning task.
     logger : logging.Logger
@@ -99,24 +108,53 @@ def __init__(
         mod: Optional[IRModule] = None,
         *,
         target: Optional[Target] = None,
-        space_generator: Optional["SpaceGenerator"] = None,
-        search_strategy: Optional["SearchStrategy"] = None,
-        sch_rules: Optional[List["ScheduleRule"]] = None,
-        postprocs: Optional[List["Postproc"]] = None,
-        mutator_probs: Optional[Dict["Mutator", float]] = None,
+        space_generator: Union[None, "SCH_FN_TYPE", "ScheduleFn", "SpaceGenerator"] = None,
+        search_strategy: Union[None, "SearchStrategy", "TuneConfig"] = None,
+        sch_rules: Union[None, str, List["ScheduleRule"]] = None,
+        postprocs: Union[None, str, List["Postproc"]] = None,
+        mutator_probs: Union[None, str, Dict["Mutator", float]] = None,
         task_name: str = "main",
         logger: Optional[logging.Logger] = None,
         rand_state: int = -1,
         num_threads: Optional[int] = None,
     ):
+        # pylint: disable=import-outside-toplevel
+        from . import default_config
+        from .space_generator import ScheduleFn
+        from .tune import TuneConfig
+
+        # pylint: enable=import-outside-toplevel
         if isinstance(mod, PrimFunc):
             mod = IRModule.from_expr(mod)
-        if num_threads is None:
-            num_threads = cpu_count()
+        if callable(space_generator):
+            space_generator = ScheduleFn(space_generator)
+        if isinstance(search_strategy, TuneConfig):
+            search_strategy = search_strategy.create_strategy()
+        if isinstance(sch_rules, str):
+            if sch_rules == "default":
+                if target is None:
+                    raise ValueError("target is required when sch_rules is 'default'")
+                sch_rules = default_config.schedule_rules(None, target)
+            else:
+                raise ValueError("sch_rules should be a list of ScheduleRule or 'default'")
+        if isinstance(postprocs, str):
+            if postprocs == "default":
+                if target is None:
+                    raise ValueError("target is required when postprocs is 'default'")
+                postprocs = default_config.postproc(None, target)
+            else:
+                raise ValueError("postprocs should be a list of Postproc or 'default'")
+        if isinstance(mutator_probs, str):
+            if mutator_probs == "default":
+                if target is None:
+                    raise ValueError("target is required when mutator_probs is 'default'")
+                mutator_probs = default_config.mutator_probs(None, target)
         if logger is None:
             self.logger = logging.getLogger(__name__)
         else:
             self.logger = None
+        if num_threads is None:
+            num_threads = cpu_count()
         self.__init_handle_by_constructor__(
             _ffi_api.TuneContext,  # type: ignore # pylint: disable=no-member
             mod,
@@ -131,9 +169,6 @@ def __init__(
             rand_state,
             num_threads,
         )
-
-    def initialize(self):
-        """Initialize the tuning context"""
         _ffi_api.TuneContextInitialize(self)  # type: ignore # pylint: disable=no-member
 
     def generate_design_space(self) -> List[Schedule]:
@@ -157,7 +192,7 @@ def generate_design_space(self) -> List[Schedule]:
 
     def pre_tuning(
         self,
-        design_spaces: List[Schedule],
+        design_spaces: Optional[List[Schedule]] = None,
         database: Optional["Database"] = None,
         cost_model: Optional["CostModel"] = None,
     ) -> None:
@@ -167,18 +202,38 @@ def pre_tuning(
 
         Parameters
         ----------
-        design_spaces : List[Schedule]
+        design_spaces : Optional[List[Schedule]]
             The design spaces used during tuning process.
+            If None, use the outcome of `self.generate_design_space()`.
         database : Optional[Database] = None
             The database used during tuning process.
+            If None, and the search strategy is `EvolutionarySearch`,
+            then use `tvm.meta_schedule.database.MemoryDatabase`.
         cost_model : Optional[CostModel] = None
             The cost model used during tuning process.
+            If None, and the search strategy is `EvolutionarySearch`,
+            then use `tvm.meta_schedule.cost_model.RandomModel`.
         """
+        # pylint: disable=import-outside-toplevel
+        from .cost_model import RandomModel
+        from .database import MemoryDatabase
+        from .search_strategy import EvolutionarySearch
+
+        # pylint: enable=import-outside-toplevel
+
         if self.search_strategy is None:
             raise ValueError(
                 "search_strategy is not provided."
                 "Please construct TuneContext with search_strategy"
             )
+        if design_spaces is None:
+            design_spaces = self.generate_design_space()
+        if database is None:
+            if isinstance(self.search_strategy, EvolutionarySearch):
+                database = MemoryDatabase()  # type: ignore
+        if cost_model is None:
+            if isinstance(self.search_strategy, EvolutionarySearch):
+                cost_model = RandomModel()  # type: ignore
         return self.search_strategy.pre_tuning(design_spaces, database, cost_model)
 
     def post_tuning(self) -> None:
@@ -191,7 +246,7 @@ def post_tuning(self) -> None:
                 "search_strategy is not provided."
                 "Please construct TuneContext with search_strategy"
             )
-        _ffi_api.SearchStrategyPostTuning(self)  # type: ignore # pylint: disable=no-member
+        return self.search_strategy.post_tuning()
 
     def generate_measure_candidates(self) -> Optional[List["MeasureCandidate"]]:
         """Generate a batch of measure candidates from design spaces for measurement.
@@ -208,7 +263,7 @@ def generate_measure_candidates(self) -> Optional[List["MeasureCandidate"]]:
                 "search_strategy is not provided."
                 "Please construct TuneContext with search_strategy"
             )
-        return _ffi_api.SearchStrategyGenerateMeasureCandidates(self)  # type: ignore # pylint: disable=no-member
+        return self.search_strategy.generate_measure_candidates()
 
     def notify_runner_results(
         self,
@@ -231,8 +286,4 @@ def notify_runner_results(
                 "search_strategy is not provided."
                 "Please construct TuneContext with search_strategy"
             )
-        _ffi_api.SearchStrategyNotifyRunnerResults(  # type: ignore # pylint: disable=no-member
-            self,
-            measure_candidates,
-            results,
-        )
+        return self.search_strategy.notify_runner_results(measure_candidates, results)

tests/python/unittest/test_meta_schedule_custom_rule_winograd_cpu.py

@@ -173,7 +173,6 @@ def test_conv2d_winograd_cpu():
             target,
         ),
     )
-    context.initialize()
     post_order_apply = context.space_generator
     (sch,) = post_order_apply.generate_design_space(mod)
     decisions = dict(

tests/python/unittest/test_meta_schedule_custom_rule_winograd_cuda.py

@@ -290,7 +290,6 @@ def test_conv2d_winograd_cuda():
             None, Target("cuda")
         ),
     )
-    context.initialize()
     post_order_apply = context.space_generator
     (sch,) = post_order_apply.generate_design_space(mod)
     decisions = dict(

tests/python/unittest/test_meta_schedule_integration.py

@@ -221,7 +221,7 @@ def traverse(t):
         mod,
         target="llvm",
         params=params,
-        filter_func=filter_func,
+        te_filter_func=filter_func,
     )
     expected_task_names = [
         "fused_" + s

tests/python/unittest/test_meta_schedule_mutator_mutate_compute_location.py

@@ -69,7 +69,6 @@ def _make_mutator(target: Target) -> Mutator:
             MutateComputeLocation(): 1.0,
         },
     )
-    ctx.initialize()
     return list(ctx.mutator_probs.keys())[0]
 
 

tests/python/unittest/test_meta_schedule_mutator_mutate_parallel.py

@@ -87,7 +87,6 @@ def _make_mutator(target: Target, max_jobs_per_core: int) -> Mutator:
             MutateParallel(max_jobs_per_core): 1.0,
         },
     )
-    ctx.initialize()
     return list(ctx.mutator_probs.keys())[0]
 
 

tests/python/unittest/test_meta_schedule_mutator_mutate_thread_binding.py

@@ -70,7 +70,6 @@ def _make_mutator(target: Target) -> Mutator:
             MutateThreadBinding(): 1.0,
         },
     )
-    ctx.initialize()
     return list(ctx.mutator_probs.keys())[0]
 
 

tests/python/unittest/test_meta_schedule_mutator_mutate_tile_size.py

@@ -73,7 +73,6 @@ def _make_mutator(target: Target) -> Mutator:
         target=target,
         mutator_probs={MutateTileSize(): 1.0},
     )
-    ctx.initialize()
     return list(ctx.mutator_probs.keys())[0]
 
 

tests/python/unittest/test_meta_schedule_mutator_mutate_unroll.py

@@ -92,7 +92,6 @@ def _make_mutator(target: Target) -> Mutator:
             MutateUnroll(): 1.0,
         },
     )
-    ctx.initialize()
     return list(ctx.mutator_probs.keys())[0]
 
 

tests/python/unittest/test_meta_schedule_post_order_apply.py

@@ -223,7 +223,6 @@ def test_meta_schedule_post_order_apply():
         space_generator=PostOrderApply(),
         sch_rules=[WowSoFancyScheduleRule()],
     )
-    context.initialize()
     post_order_apply = context.space_generator
     schs = post_order_apply.generate_design_space(mod)
     assert len(schs) == 1
@@ -240,7 +239,6 @@ def test_meta_schedule_post_order_apply_double():
         space_generator=PostOrderApply(),
         sch_rules=[DoubleScheduleRule()],
     )
-    context.initialize()
     post_order_apply = context.space_generator
     schs = post_order_apply.generate_design_space(mod)
     assert len(schs) == 2
@@ -258,7 +256,6 @@ def test_meta_schedule_post_order_apply_multiple():
         space_generator=PostOrderApply(),
         sch_rules=[DoubleScheduleRule(), ReorderScheduleRule()],
     )
-    context.initialize()
     post_order_apply = context.space_generator
     schs = post_order_apply.generate_design_space(mod)
     assert len(schs) == 4
@@ -276,7 +273,6 @@ def test_meta_schedule_post_order_apply_duplicate_matmul():
         space_generator=PostOrderApply(),
         sch_rules=[WowSoFancyScheduleRule()],
     )
-    context.initialize()
     post_order_apply = context.space_generator
     with pytest.raises(
         TVMError,
@@ -348,7 +344,6 @@ def correct_trace(a, b, c, d):
         space_generator=PostOrderApply(),
         sch_rules=[RemoveBlock(), TrinityDouble()],
     )
-    context.initialize()
     post_order_apply = context.space_generator
     schs = post_order_apply.generate_design_space(mod)
     assert len(schs) == 4
@@ -376,7 +371,6 @@ def test_meta_schedule_custom_search_space():
         space_generator=PostOrderApply(),
         sch_rules=[],
     )
-    context.initialize()
     post_order_apply = context.space_generator
     post_order_apply.generate_design_space(mod)
     called = False

tests/python/unittest/test_meta_schedule_postproc_disallow_dynamic_loop.py

@@ -37,7 +37,6 @@ def _create_context(mod, target) -> TuneContext:
         ],
         task_name="test",
     )
-    ctx.initialize()
     return ctx
 
 

tests/python/unittest/test_meta_schedule_postproc_rewrite_cooperative_fetch.py

@@ -39,7 +39,6 @@ def _create_context(mod, target) -> TuneContext:
         ],
         task_name="test",
     )
-    ctx.initialize()
     return ctx
 
 

tests/python/unittest/test_meta_schedule_postproc_rewrite_reduction_block.py

@@ -37,7 +37,6 @@ def _create_context(mod, target) -> TuneContext:
         ],
         task_name="test",
     )
-    ctx.initialize()
     return ctx
 
 

tests/python/unittest/test_meta_schedule_postproc_rewrite_tensorize.py

@@ -457,7 +457,6 @@ def _create_context(mod, target, postprocs):
         postprocs=postprocs,
         task_name="test",
     )
-    ctx.initialize()
     return ctx