Minor fix.

python/tvm/meta_schedule/task_scheduler/gradient_based.py

@@ -36,15 +36,13 @@
 
 
 @register_func("meta_schedule.task_scheduler.derive_similarity_tag")
-def derive_similarity_tag(mod: IRModule, log_base: float = 1.618) -> str:
+def derive_similarity_tag(mod: IRModule) -> str:
     """Get the tags for smilarity group creation
 
     Parameters
     ----------
     mod : IRModule
         The input workload.
-    log_base : float
-        The log base to normalize the flop count. Default natural (1.618).
 
     Return
     ------
@@ -57,7 +55,7 @@ def derive_similarity_tag(mod: IRModule, log_base: float = 1.618) -> str:
             ret += mod[var].attrs.meta_scheduler_task_scheduler_tag + "_"
     if ret:
         flop_count = _ffi_api.TaskSchedulerFlopCount(mod)  # type: ignore # pylint: disable=no-member
-        ret += "%d" % int(math.log(flop_count + 1, log_base))
+        ret += "%d" % int(math.log(flop_count + 1, 1.618))
     return ret
 
 

src/meta_schedule/task_scheduler/gradient_based.cc

@@ -79,13 +79,13 @@ class GradientBasedNode final : public TaskSchedulerNode {
     for (int i : task_groups[group_id]) {
       best_flops = std::max(best_flops, task_flop_counts[i] / task_best_latencies[i]);
       max_cnt[0] = task_cnts[i];
-      std::sort(max_cnt, max_cnt + 3);  // place the 2nd largest to #1
+      std::sort(max_cnt, max_cnt + 3);  // place the 2nd largest to middle position
     }
     double cur_flops = task_flop_counts[task_id] / task_best_latencies[task_id];
     // if we tune a task for many times but it still cannot achieve
     // a similar speed to the fastest one in its group, this means this task
     // is actually not similar to other tasks in its group.
-    // So we will remove it from its original group.
+    // So we will move it from the current group to a standalone group.
 
     if (cur_flops < best_flops / beta && task_cnts[task_id] > 5 + max_cnt[1]) {
       task_groups[group_id].erase(task_id);