[Minor][Inductor] move some combo kernel log from warning to debug (pytorch#167020)

[Minor][Inductor] move some combo kernel log from warning to debug (pytorch#166993)

Combo kernel warns for long reduction and large pointwise. This becomes too spammy for users such as vLLM.

This PR moves these logs from warn to debug. I validated the spammy log is removed on llama-3.1-8B.

Pull Request resolved: pytorch#166993
Approved by: https://github.com/zou3519, https://github.com/eellison

(cherry picked from commit e020fb3)

Co-authored-by: Boyuan Feng <[email protected]>

Author: pytorchbot

torch/_inductor/codegen/triton_combo_kernel.py

@@ -94,7 +94,7 @@ def _default_custom_combo_kernel_horizontal_partition(
         ]
         short_reduction = [n for n in reduction if n not in long_reduction]
         if long_reduction:
-            log.warning(
+            log.debug(
                 "ComboKernels: %d long reduction nodes are separated",
                 len(long_reduction),
             )
@@ -107,7 +107,7 @@ def _default_custom_combo_kernel_horizontal_partition(
         ]
         if large_pointwise:
             # TODO benchmark the performance when large pointwise nodes combining with others
-            log.warning(
+            log.debug(
                 "ComboKernels: %d large pointwise nodes are separated",
                 len(large_pointwise),
             )