[AutoScheduler] Support early_stopping per task (#7377)

* [AutoScheduler] Support early_stopping per task

* address comment

* fix test

* Update python/tvm/auto_scheduler/task_scheduler.py

* Update python/tvm/auto_scheduler/task_scheduler.py

* trigger ci

* trigger ci

Author: comaniac

python/tvm/auto_scheduler/task_scheduler.py

@@ -246,6 +246,9 @@ def __init__(
         # task_cts[i] saves how many times task i is tuned
         self.task_cts = [0 for _ in range(len(self.tasks))]
 
+        # task_best_cts[i] saves the round task i found the best latency
+        self.task_best_cts = [0 for _ in range(len(self.tasks))]
+
         # task_costs_history[i] saves the latency history of task i
         self.task_costs_history = [[] for _ in range(len(self.tasks))]
 
@@ -281,13 +284,14 @@ def tune(
         search_policy="default",
         search_policy_params=None,
         adapative_training=False,
+        per_task_early_stopping=None,
     ):
         """Tune a batch of tasks together.
 
         Parameters
         ----------
         tune_option: TuningOptions
-            The options of tuning
+            The tuning options applied to all tasks.
         search_policy: : Union[str, List[SearchPolicy]] = "default"
             The list of search policies.
             If it is str,
@@ -299,10 +303,17 @@ def tune(
         adapative_training : bool = False
             Option used by XGBModel to reduce the model training frequency when there're
             too many logs.
+        per_task_early_stopping : Optional[int]
+            Stop tuning a task early if getting no improvement after n measurements.
         """
         # init members
         self.tune_option = tune_option
-        early_stopping = 1e20 if tune_option.early_stopping < 0 else tune_option.early_stopping
+        self.early_stopping_all = (
+            1e20 if tune_option.early_stopping < 0 else tune_option.early_stopping
+        )
+        self.early_stopping_task = (
+            1e20 if per_task_early_stopping is None else per_task_early_stopping
+        )
 
         self.measurer = ProgramMeasurer(
             tune_option.builder,
@@ -417,13 +428,13 @@ def tune(
             if self.cur_score < self.best_score:
                 self.best_score = self.cur_score
                 self.best_ct = self.ct
-            elif self.ct - self.best_ct >= early_stopping and all(
+            elif self.ct - self.best_ct >= self.early_stopping_all and all(
                 cost < 1e9 for cost in self.best_costs
             ):
                 if self.tune_option.verbose >= 1:
                     print(
                         "Stop early since no performance improvement in the last "
-                        + str(early_stopping)
+                        + str(self.early_stopping_all)
                         + " measurement trials."
                     )
                 break
@@ -439,15 +450,22 @@ def _tune_task(self, task_idx):
             self.num_measures_per_round, self.measurer
         )
 
+        self.task_cts[task_idx] += 1
+
         for res in measure_results:
             cost = array_mean(res.costs)
             if cost < self.best_costs[task_idx]:
+                self.task_best_cts[task_idx] = self.task_cts[task_idx]
                 self.best_costs[task_idx] = cost
 
-        if len(measure_inputs) == 0:
+        # Stop tuning this task in the rest of the process if its search space has been
+        # fully explored or it has no improvement for a long while.
+        no_change_trials = (
+            self.task_cts[task_idx] - self.task_best_cts[task_idx]
+        ) * self.num_measures_per_round
+        if len(measure_inputs) == 0 or no_change_trials > self.early_stopping_task:
             self.dead_tasks.add(task_idx)
 
-        self.task_cts[task_idx] += 1
         self.task_costs_history[task_idx].append(self.best_costs[task_idx])
 
         self.ct += len(measure_inputs)
@@ -494,17 +512,24 @@ def _restore_status(self, log_file, num_measures_per_round):
             if task_idx is None:
                 continue
 
+            self.task_cts[task_idx] += 1
+
             if res.error_no == 0:
-                self.best_costs[task_idx] = min(self.best_costs[task_idx], array_mean(res.costs))
+                cost = array_mean(res.costs)
+                if self.best_costs[task_idx] < cost:
+                    self.best_costs[task_idx] = cost
+                    self.task_best_cts = self.task_cts[task_idx]
 
-            self.task_cts[task_idx] += 1
+        for idx in range(len(self.tasks)):
+            if self.task_cts[idx] - self.task_best_cts[idx] > self.early_stopping_task:
+                self.dead_tasks.add(idx)
 
-        for i in range(len(self.tasks)):
             # The computation of taks_cts is just an estimation.
             # The estimation may not be accurate if the log file is changed externally or
             # `num_measures_per_round` is different from the last tuning.
-            self.task_cts[i] = int(self.task_cts[i] / num_measures_per_round + 0.5)
-            self.task_costs_history[i].append(self.best_costs[i])
+            self.task_cts[idx] = int(self.task_cts[idx] / num_measures_per_round + 0.5)
+            self.task_best_cts[idx] = int(self.task_best_cts[idx] / num_measures_per_round + 0.5)
+            self.task_costs_history[idx].append(self.best_costs[idx])
 
         self.cur_score = self._compute_score(self.best_costs)