[TVMC] A simplified TVMC API for python scripting (Part 1). (#7823)

* Introduce new TVMC Python API.

* Add simple testing model.

* Split result utils into stand-alone file.

python/tvm/auto_scheduler/search_task.py

@@ -43,40 +43,74 @@
 
 @tvm._ffi.register_object("auto_scheduler.HardwareParams")
 class HardwareParams(Object):
-    """The parameters of target hardware used to guide the search policy
+    """The parameters of target hardware used to guide the search policy.
+
+    When a parameter isn't provided, it will instead use the
+    current machine's default value if target is specified.
     TODO(jcf94): This is considered to be merged with the new Target specification:
     https://discuss.tvm.apache.org/t/rfc-tvm-target-specification/6844
     Parameters
     ----------
-    num_cores : int
+    num_cores : int, optional
         The number of device cores.
-    vector_unit_bytes : int
+    vector_unit_bytes : int, optional
         The width of vector units in bytes.
-    cache_line_bytes : int
+    cache_line_bytes : int, optional
         The size of cache line in bytes.
-    max_shared_memory_per_block : int
+    max_shared_memory_per_block : int, optional
         The max shared memory per block in bytes.
-    max_local_memory_per_block : int
+    max_local_memory_per_block : int, optional
         The max local memory per block in bytes.
-    max_threads_per_block : int
+    max_threads_per_block : int, optional
         The max number of threads per block.
-    max_vthread_extent : int
+    max_vthread_extent : int, optional
         The max vthread extent.
-    warp_size : int
+    warp_size : int, optional
         The thread numbers of a warp.
+    target : str or Target, optional
+        The compilation target. Used to determine default values if provided.
+    target_host : str or Target, optional
+        The compilation target host. Used to determine default values if provided.
     """
 
     def __init__(
         self,
-        num_cores,
-        vector_unit_bytes,
-        cache_line_bytes,
-        max_shared_memory_per_block,
-        max_local_memory_per_block,
-        max_threads_per_block,
-        max_vthread_extent,
-        warp_size,
+        num_cores=None,
+        vector_unit_bytes=None,
+        cache_line_bytes=None,
+        max_shared_memory_per_block=None,
+        max_local_memory_per_block=None,
+        max_threads_per_block=None,
+        max_vthread_extent=None,
+        warp_size=None,
+        target=None,
+        target_host=None,
     ):
+        # If target is provided, get the default paramters for this machine.
+        if target is not None:
+            if isinstance(target, str):
+                target = tvm.target.Target(target)
+            if isinstance(target_host, str):
+                target_host = tvm.target.Target(target_host)
+            default_params = _ffi_api.GetDefaultHardwareParams(target, target_host)
+
+            if num_cores is None:
+                num_cores = default_params.num_cores
+            if vector_unit_bytes is None:
+                vector_unit_bytes = default_params.vector_unit_bytes
+            if cache_line_bytes is None:
+                cache_line_bytes = default_params.cache_line_bytes
+            if max_shared_memory_per_block is None:
+                max_shared_memory_per_block = default_params.max_shared_memory_per_block
+            if max_local_memory_per_block is None:
+                max_local_memory_per_block = default_params.max_local_memory_per_block
+            if max_threads_per_block is None:
+                max_threads_per_block = default_params.max_threads_per_block
+            if max_vthread_extent is None:
+                max_vthread_extent = default_params.max_vthread_extent
+            if warp_size is None:
+                warp_size = default_params.warp_size
+
         self.__init_handle_by_constructor__(
             _ffi_api.HardwareParams,
             num_cores,
@@ -89,6 +123,21 @@ def __init__(
             warp_size,
         )
 
+    def __str__(self):
+        """Pretty printing for hardware parameter configuration."""
+        format_str = (
+            "HardwareParams:\n"
+            f"  num_cores: {self.num_cores}\n"
+            f"  vector_unit_bytes: {self.vector_unit_bytes}\n"
+            f"  cache_line_bytes: {self.cache_line_bytes}\n"
+            f"  max_shared_memory_per_block: {self.max_shared_memory_per_block}\n"
+            f"  max_local_memory_per_block: {self.max_local_memory_per_block}\n"
+            f"  max_threads_per_block: {self.max_threads_per_block}\n"
+            f"  max_vthread_extent: {self.max_vthread_extent}\n"
+            f"  warp_size: {self.warp_size}\n"
+        )
+        return format_str
+
 
 @tvm._ffi.register_object("auto_scheduler.TuningOptions")
 class TuningOptions(Object):

python/tvm/auto_scheduler/task_scheduler.py

@@ -329,7 +329,10 @@ def tune(
             tune_option.num_measures_per_round, tune_option.num_measure_trials // len(self.tasks)
         )
         if self.num_measures_per_round <= 0:
-            raise ValueError("num_measure_trials is too small. Please set it to a higher value.")
+            raise ValueError(
+                "num_measure_trials is too small. Please set it to a higher value."
+                f"It should be at least {len(self.tasks)} for this model."
+            )
 
         # restore the status of the task scheduler from a log file
         if self.load_log_file:

python/tvm/driver/tvmc/__init__.py

@@ -22,6 +22,9 @@
 from . import autotuner
 from . import compiler
 from . import runner
+from . import result_utils
 from .frontends import load_model as load
 from .compiler import compile_model as compile
 from .runner import run_module as run
+from .autotuner import tune_model as tune
+from .model import TVMCModel, TVMCPackage, TVMCResult