Skip to content
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.

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

Override estimated speedups when estimation model is enabled #885

Merged
merged 4 commits into from
Mar 27, 2024

Conversation

parthosa
Copy link
Collaborator

@parthosa parthosa commented Mar 26, 2024

Fixes #878. This PR introduces the following changes:

  1. Updates the user tools to replace the Estimated GPU Speedup by the speed ups generated from QualX.
    • As a part of this, add a new column Speed Up Estimation Model which stores which estimation model was used for calculation of speed up (XGBoost or Speedup (heuristic approach)).
    • This column should be present even if the xgboost estimation was disabled.
  2. Modify the output directory from QualX to xgboost_predictions and output files as per_sql.csv and per_app.csv.
  3. Updates the prediction function to use model based on the platform.
  4. Introduce debug variable to toggle printing debug information. Default is False

Output

Platform: onprem:

qualification_summary.csv
,App Name,App ID,Speedup Based Recommendation,Estimated GPU Speedup,Estimated GPU Duration,App Duration,Unsupported Operators Stage Duration,Unsupported Operators Stage Duration Percent,Speed Up Estimation Model
0,NDS - query72,application_1686676198636_0002,Recommended,7.04,126117.00,887621.00,0.00,0.00,XGBOOST
1,Databricks Shell,app-20230906000351-0000,Recommended,6.45,788394.00,5088279.00,83674.00,1.64,XGBOOST
Report Summary:
-------------------------  ----
Total applications            2
RAPIDS candidates             2
Overall estimated speedup  6.53
-------------------------  ----

Platform: dataproc

qualification_summary.csv
,App Name,App ID,Speedup Based Recommendation,Estimated GPU Speedup,Estimated GPU Duration,App Duration,Unsupported Operators Stage Duration,Unsupported Operators Stage Duration Percent,Speed Up Estimation Model
0,NDS - query72,application_1686676198636_0002,Strongly Recommended,4.92,180511.00,887621.00,0.00,0.00,XGBOOST
1,Databricks Shell,app-20230906000351-0000,Recommended,3.91,1302505.00,5088279.00,83674.00,1.64,XGBOOST
Report Summary:
-------------------------  ----
Total applications            2
RAPIDS candidates             2
Overall estimated speedup  4.03
-------------------------  ----

Platform: onprem with XG boost estimation disabled

qualification_summary.csv
,App Name,App ID,Speedup Based Recommendation,Estimated GPU Speedup,Estimated GPU Duration,App Duration,Unsupported Operators Stage Duration,Unsupported Operators Stage Duration Percent,Speed Up Estimation Model
0,NDS - query72,application_1686676198636_0002,Strongly Recommended,3.28,270531.04,887621.00,0.00,0.00,SPEEDUPS
1,Databricks Shell,app-20230906000351-0000,Recommended,1.78,2865463.88,5088279.00,83674.00,1.64,SPEEDUPS
Report Summary:
-------------------------  ----
Total applications            2
RAPIDS candidates             2
Overall estimated speedup  1.91
-------------------------  ----

…te model based on platform

Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
@parthosa parthosa added feature request New feature or request user_tools Scope the wrapper module running CSP, QualX, and reports (python) labels Mar 26, 2024
@parthosa parthosa self-assigned this Mar 26, 2024
Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
Copy link
Collaborator

@amahussein amahussein left a comment

Choose a reason for hiding this comment

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

Thanks @parthosa
Everything looks good for first iteration. I have quick suggestion though.

Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
Signed-off-by: Partho Sarthi <psarthi@nvidia.com>
Copy link
Collaborator

@cindyyuanjiang cindyyuanjiang left a comment

Choose a reason for hiding this comment

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

Thanks @parthosa!

Copy link
Collaborator

@amahussein amahussein left a comment

Choose a reason for hiding this comment

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

Thanks @parthosa

@amahussein amahussein merged commit cfed5c0 into NVIDIA:dev Mar 27, 2024
15 checks passed
@parthosa parthosa deleted the spark-rapids-tools-878 branch March 27, 2024 15:51
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
feature request New feature or request user_tools Scope the wrapper module running CSP, QualX, and reports (python)
Projects
None yet
Development

Successfully merging this pull request may close these issues.

[FEA] Override Estimated speedups when estimation model is enabled
3 participants