Update

ezyang · ezyang · commit df08081f87c8 · 2024-11-26T19:07:39.000-08:00
[ghstack-poisoned]
diff --git a/torch/_inductor/codecache.py b/torch/_inductor/codecache.py
@@ -1638,6 +1638,10 @@ def load(  # type: ignore[no-untyped-def]
         # with log_cache_bypass.
         chromium_log = get_chromium_event_logger()
         cache_state = cache_info["cache_state"]
+        # Here for grepping:
+        # fx_graph_cache_hit
+        # fx_graph_cache_miss
+        # fx_graph_cache_bypass
         chromium_log.log_instant_event(
             f"fx_graph_cache_{cache_state}",
             cache_info["cache_event_time"],
diff --git a/torch/_inductor/compile_fx.py b/torch/_inductor/compile_fx.py
@@ -1446,7 +1446,7 @@ def compile_fx(
     decompositions: Optional[Dict[OpOverload, Callable[..., Any]]] = None,
 ) -> Union[Callable[[List[object]], Sequence[torch.Tensor]], str, List[str]]:
     """
-    Main entrypoint to a compile given FX graph.  Despite the fact that this
+    Main entry point for compiling given FX graph.  Despite the fact that this
     lives in :mod:`torch._inductor`, this function is responsible for calling
     into AOT Autograd (and we will eventually get a callback to
     ``inner_compile`` to perform actual compilation.  In other words, this
@@ -1637,7 +1637,7 @@ def fw_compiler_base(
                     orig_output_end_idx = (
                         original_output_start_index + num_orig_model_outputs
                     )
-                    # Sanity chec: we are about to splice out the "user" outputs from the full set
+                    # Sanity check: we are about to splice out the "user" outputs from the full set
                     # of "graph" outputs. Make sure we're within bounds.
                     assert orig_output_end_idx <= num_model_outputs
 
