frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 21:14:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

Don't upload compiler benchmark debug info to the benchmark database (#153769)

Browse Source 
During our debug session, @wdvr and I found out that the benchmark database is growing much faster than we expect.  After taking a closer look, the majority of them coming from TorchInductor benchmark and the top 3 are all debug information not used by any dashboard atm.  In the period of 7 days, there are close to 6 millions records ([query](https://paste.sh/GUVCBa0v#UzszFCZaWQxh7oSVsZtfZdVE))

```
Benchmark,Metric,Count
"TorchInductor","user_stack","1926014"
"TorchInductor","reason","1926014"
"TorchInductor","model","1926014"
```

Let's skip uploading them to avoid bloating the database.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/153769
Approved by: https://github.com/malfet
This commit is contained in:
Huy Do
2025-05-23 01:18:26 +00:00
committed by PyTorch MergeBot
parent 768cb734ec
commit 7509b150af
1 changed files with 8 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
benchmarks/dynamo/common.py
View File

@ -343,7 +343,7 @@ def load_model_from_path(path_and_class_str):
return model, inputs
def write_outputs(filename, headers, row):
def write_outputs(filename, headers, row, upload_to_benchmark_db: bool = True):
"""
Write both CSV and JSON outputs using the original CSV output interface
"""
@ -352,7 +352,8 @@ def write_outputs(filename, headers, row):
return
output_csv(filename, headers, row)
output_json(filename, headers, row)
if upload_to_benchmark_db:
output_json(filename, headers, row)
def output_csv(filename, headers, row):
@ -2847,10 +2848,15 @@ class BenchmarkRunner:
user_stack = add_double_quotes(
", ".join([str(x) for x in graph_break.user_stack])
)
# NB: Don't upload them to the benchmark database as they are debugging
# infomation. There are also around a million records a day which is
# wasteful to store
write_outputs(
filename,
["model", "reason", "user_stack"],
[current_name, reason, user_stack],
False,
)
if self.args.stats: