Cherry-pick #4168 to main: Fix hello-numpy-cross-val example

[nvkevlu]

Cherry-pick #4168 to main: Fix hello-numpy-cross-val example.

Description

Cherry-pick #4168 to main: Fix hello-numpy-cross-val example.
(#4011 and #4173 already on main; only 4168 changes applied.)

Types of changes

• Non-breaking change (fix or new feature that would not break existing functionality).
• Breaking change (fix or new feature that would cause existing functionality to change).
• New tests added to cover the changes.
• Quick tests passed locally by running ./runtest.sh.
• In-line docstrings updated.
• Documentation updated.

[nvkevlu]

/build

[greptile-apps]

Greptile Overview

Greptile Summary

Cherry-picked PR #4168 to fix the hello-numpy-cross-val example by refactoring the client to properly support cross-site evaluation (CSE) workflow with distinct task handling.

Key Changes:

• Refactored client.py to handle three task types separately: train, evaluate, and submit_model with explicit branching
• Added fail-fast error handling for train tasks with missing params to prevent empty responses that would break server aggregation (aligns with custom rule 783565ac-d530-4d49-a8bc-55877cb0a0cd)
• Introduced last_params tracking to store trained weights for submission during CSE submit_model phase
• Fixed integration test config key from class_path to path for dict-based model configuration
• Added comprehensive unit tests verifying helper functions and fail-fast behavior for invalid scenarios

Confidence Score: 5/5

• This PR is safe to merge with minimal risk
• Cherry-picked fix with clear improvements to error handling, comprehensive test coverage, and proper task separation. Changes follow fail-fast principles per custom instructions, prevent silent failures that could break aggregation, and include both unit and integration tests
• No files require special attention

Important Files Changed

Filename	Overview
examples/hello-world/hello-numpy-cross-val/client.py	Refactored client to handle multiple task types (train, evaluate, submit_model) with explicit fail-fast error handling for missing parameters, tracking last_params for CSE workflow
tests/integration_test/recipe_system_test.py	Fixed model config key from class_path to path for consistency with dict-based config format
tests/unit_test/recipe/hello_numpy_cross_val_client_test.py	New comprehensive unit tests for hello-numpy-cross-val client covering helper functions and fail-fast behavior for invalid train/submit_model scenarios

Flowchart

flowchart TD
    Start([Client receives task]) --> CheckType{Task Type?}
    
    CheckType -->|is_train| ValidateParams{Params valid?}
    ValidateParams -->|No params or missing NUMPY_KEY| RaiseError1[Raise RuntimeError: Train task needs params]
    ValidateParams -->|Valid| Train[Train model: params + 1]
    Train --> SaveParams[Store in last_params]
    SaveParams --> Evaluate1[Evaluate model]
    Evaluate1 --> CheckUpdateType{update_type?}
    CheckUpdateType -->|diff| SendDiff[Send params diff]
    CheckUpdateType -->|full| SendFull[Send full params]
    SendDiff --> End([Continue loop])
    SendFull --> End
    
    CheckType -->|is_evaluate| ValidateEvalParams{Params valid?}
    ValidateEvalParams -->|No| SendEmpty1[Send empty metrics]
    ValidateEvalParams -->|Yes| Evaluate2[Evaluate model]
    Evaluate2 --> SendMetrics[Send metrics only]
    SendMetrics --> End
    SendEmpty1 --> End
    
    CheckType -->|is_submit_model| CheckLastParams{last_params set?}
    CheckLastParams -->|No| RaiseError2[Raise RuntimeError: No local model]
    CheckLastParams -->|Yes| SubmitModel[Submit last_params as FULL]
    SubmitModel --> End
    
    CheckType -->|Other| SendEmpty2[Send empty metrics]
    SendEmpty2 --> End
    
    RaiseError1 --> Stop([Fail fast])
    RaiseError2 --> Stop

Loading

_{Last reviewed commit: 72120c5}

[greptile-apps]

_{3 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}

[Copilot]

Pull request overview

This PR cherry-picks PR #4168 to the main branch, fixing the hello-numpy-cross-val example to properly support cross-site evaluation (CSE) by implementing branching logic for different task types.

Changes:

• Refactored hello-numpy-cross-val client to use branching pattern with is_train(), is_evaluate(), and is_submit_model() instead of assuming all tasks are training tasks
• Added fail-fast validation that raises RuntimeError when train tasks receive no model params or when submit_model is called without trained params
• Added comprehensive unit tests for the client's helper functions and fail-fast behavior
• Fixed integration test that incorrectly used "class_path" instead of "path" for model configuration dict

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
tests/unit_test/recipe/hello_numpy_cross_val_client_test.py	New unit test file that validates client helper functions (train/evaluate) and tests fail-fast behavior when params are missing
tests/integration_test/recipe_system_test.py	Fixed model config dict to use "path" instead of "class_path" (correct for job/config layer)
examples/hello-world/hello-numpy-cross-val/client.py	Refactored to use branching pattern for train/evaluate/submit_model tasks with proper error handling and last_params tracking for CSE

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[greptile-apps]

_{3 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}

[nvkevlu]

/build

[nvkevlu]

/build

[greptile-apps]

_{3 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}