Introduced TORCH_TRACE/tlparse

Author: nluu175

recipes_source/torch_logs.py

@@ -88,8 +88,24 @@ def fn(x, y):
         except Exception as e:
             print(f"Error during {name}: {str(e)}")
 
-if __name__ == "__main__":
-    run_debugging_suite()
+
+run_debugging_suite()
+
+######################################################################
+# Use TORCH_TRACE/tlparse to produce produce compilation reports
+# ~~~~~~~~~~
+#
+# In this section, we introduce the usage of TORCH_TRACE and tlparse to produce reports.
+# First, we run `TORCH_TRACE="/tmp/tracedir" python script.py` to generate the raw trace logs.
+# We have replace `/tmp/tracedir` with a path to a directory you want to store the trace logs
+# and reaplce script with the name of your script.
+# 
+# Next, we are going to pass the trace log to `tlparse` to generate compilation reports. We run
+# `pip install tlparse` and then `tlparse /tmp/tracedir`. This will open up your browser with 
+# HTML like generated above.
+#
+# By default, reports generated by `tlparse` are 
+# stored in the directory `tl_out`. You can change that by doing `tlparse /tmp/tracedir -o output_dir/`.
 
 ######################################################################
 # Conclusion