feat: add power estimation #153
Open
+3,148
−808
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
kaim-eng
requested review from
AichenF,
Arsene12358,
Ethan-ES,
YijiaZhao,
davilu-nvidia,
ilyasher,
jasonqinzhou,
simone-chen,
tianhaox and
xutizhou
as code owners
kaim-eng
force-pushed
the
new-104
branch
2 times, most recently
from
4e3bc8f to
ae43171
Compare
- Add PerformanceResult class: float-like object that carries both latency and power data - Add power_w column to all output DataFrames (ColumnsStatic, ColumnsAgg, ColumnsDisagg) - Extend PerfDatabase with query_*_with_power methods for power-aware queries - Update all backend classes to support power estimation queries - Modify operations to return PerformanceResult with power data - Update inference session/summary to track and aggregate power consumption - Extend CLI report tables to display power_w column in results - Maintain backward compatibility with legacy database formats (defaults to 0.0W) This feature enables power consumption estimation alongside performance metrics, allowing users to optimize for both throughput and power efficiency. Signed-off-by: Kai Ma <[email protected]>
- Fix line length violations by splitting long strings - Remove useless if-else condition in perf_database - Clean up whitespace and formatting issues - All ruff checks now pass Signed-off-by: Kai Ma <[email protected]>
Update test_data_loaders.py to check for dictionary format with 'latency' and 'power' keys instead of plain float values. This aligns tests with the power estimation feature changes to data loader functions. Signed-off-by: Kai Ma <[email protected]>
Update test_interpolation.py to handle both dictionary format (with 'latency' and 'power' keys) and legacy float format. This ensures backward compatibility while supporting the new power-aware data structure. Signed-off-by: Kai Ma <[email protected]>
Signed-off-by: Kai Ma <[email protected]>
Signed-off-by: Kai Ma <[email protected]>
Signed-off-by: Kai Ma <[email protected]>
Signed-off-by: Kai Ma <[email protected]>
Signed-off-by: Kai Ma <[email protected]>
tianhaox
reviewed
Dec 8, 2025
| self._scale_factor = scale_factor | ||
|
|
||
| def query(self, database: PerfDatabase, **kwargs): | ||
| def query(self, database: PerfDatabase, **kwargs) -> float: |
Contributor
There was a problem hiding this comment.
for return, should we indicate PerformanceResult, or just float
Author
There was a problem hiding this comment.
Yes. I thought about this for quite some time. Currently, we return Performance_Result to minimize the impact to existing code
src/aiconfigurator/sdk/performance_result.py
It currently default return to float unless we tell it to return both perf and power.
I will change this to Performance_Result to be accurate
Signed-off-by: Kai Ma <[email protected]>
Signed-off-by: Kai Ma <[email protected]>
Signed-off-by: Kai Ma <[email protected]>
Signed-off-by: Kai Ma <[email protected]>
Signed-off-by: Kai Ma <[email protected]>
Signed-off-by: Kai Ma <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This is a new start of previous PR104
#104
I've been trying to rebase the dev branch of 104 to main ToT for several days but going nowhere.
This PR is basically redo what we have for 104 but from 12/3 ToT.