Skip to content

[benchmark] Add triton version in the moe tuned config #24769

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Sep 16, 2025
Merged

[benchmark] Add triton version in the moe tuned config #24769

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
cc16128
Done
jeejeelee Sep 12, 2025
eee4788
Merge branch 'main' into add-triton-version
jeejeelee Sep 13, 2025
659d8be
Merge branch 'main' into add-triton-version
jeejeelee Sep 15, 2025
2db8231
Merge branch 'main' into add-triton-version
jeejeelee Sep 16, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion benchmarks/kernels/benchmark_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -560,7 +560,7 @@ def save_configs(
filename = os.path.join(save_dir, filename)
print(f"Writing best config to {filename}...")
with open(filename, "w") as f:
json.dump(configs, f, indent=4)
json.dump({"triton_version": triton.__version__, **configs}, f, indent=4)
f.write("\n")


Expand Down
5 changes: 4 additions & 1 deletion vllm/model_executor/layers/fused_moe/fused_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -720,7 +720,10 @@ def get_moe_configs(
logger.info("Using configuration from %s for MoE layer.",
config_file_path)
# If a configuration has been found, return it
return {int(key): val for key, val in json.load(f).items()}
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

The proposed change to explicitly pop triton_version is brittle. If other metadata is added to the configuration file in the future, this code will fail with a ValueError when trying to convert a non-integer string key to an integer.

A more robust approach is to filter for keys that are valid integers. By checking key.isdigit(), you can automatically ignore any non-numeric keys, including triton_version and any future metadata, making the code more resilient to future changes.

Suggested change
return {int(key): val for key, val in json.load(f).items()}
return {int(key): val for key, val in json.load(f).items() if key.isdigit()}

tuned_config = json.load(f)
# Delete triton_version from tuned_config
tuned_config.pop("triton_version", None)
return {int(key): val for key, val in tuned_config.items()}

# If no optimized configuration is available, we will use the default
# configuration
Expand Down
2 changes: 1 addition & 1 deletion vllm/triton_utils/importing.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,7 +68,7 @@ class TritonPlaceholder(types.ModuleType):

def __init__(self):
super().__init__("triton")
self.__version__ = "3.3.0"
self.__version__ = "3.4.0"
self.jit = self._dummy_decorator("jit")
self.autotune = self._dummy_decorator("autotune")
self.heuristics = self._dummy_decorator("heuristics")
Expand Down