[PR #1337] [MERGED] test(tinytorch): add coverage tests for Module 08 training infrastructure #8128

New Issue

GiteaMirror · 2026-04-27T17:25:52-05:00

GiteaMirror commented

2026-04-27 17:25:52 -05:00

📋 Pull Request Information

Original PR: https://github.com/harvard-edge/cs249r_book/pull/1337
Author: @Shashank-Tripathi-07
Created: 4/16/2026
Status: ✅ Merged
Merged: 4/16/2026
Merged by: @profvjreddi

Base: dev ← Head: test/training-module-coverage

📝 Commits (2)

6b42b75 test(tinytorch): add coverage tests for Module 08 training infrastructure
13972d6 fix(tests/08_training): correct scheduler lr assertion to use epoch 0 not epoch 1

📊 Changes

1 file changed (+480 additions, -0 deletions)

View changed files

➕ tinytorch/tests/08_training/test_training_coverage.py (+480 -0)

📄 Description

Summary

Six areas of Module 08 were fully implemented but had zero test coverage. This PR fills those gaps with a dedicated test_training_coverage.py.

Coverage

Area	What's tested
`CosineSchedule`	Start/mid/end values, monotonic decrease, past-end edge case, single-epoch schedule
`clip_grad_norm`	Clipping, no-clip, multi-param global norm, direction preserved, zero-grad safety, returns original norm
`save/load_checkpoint`	File created, all keys present, epoch/step/history/weights restored, training resumes correctly, creates parent dirs
`Trainer.evaluate`	Finite loss, float return types, eval mode set, weights unchanged, history recorded per call, classification accuracy (correct and wrong)
Scheduler integration	LR recorded in history, `optimizer.lr` updated to match schedule, decreases over epochs
Gradient clipping integration	Training completes without crash, weights still update, tight clip limits update magnitude
Train/eval mode switching	`model.training` correct after `train_epoch`, `evaluate`, and train/eval/train cycle

Files

File	Change
`tinytorch/tests/08_training/test_training_coverage.py`	New: 478 lines, 36 tests

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/harvard-edge/cs249r_book/pull/1337 **Author:** [@Shashank-Tripathi-07](https://github.com/Shashank-Tripathi-07) **Created:** 4/16/2026 **Status:** ✅ Merged **Merged:** 4/16/2026 **Merged by:** [@profvjreddi](https://github.com/profvjreddi) **Base:** `dev` ← **Head:** `test/training-module-coverage` --- ### 📝 Commits (2) - [`6b42b75`](https://github.com/harvard-edge/cs249r_book/commit/6b42b75b55dfbec66f575bd20f27a517f9b41f9f) test(tinytorch): add coverage tests for Module 08 training infrastructure - [`13972d6`](https://github.com/harvard-edge/cs249r_book/commit/13972d6807dda94a0dc44388d5afa24cc1a11efc) fix(tests/08_training): correct scheduler lr assertion to use epoch 0 not epoch 1 ### 📊 Changes **1 file changed** (+480 additions, -0 deletions) <details> <summary>View changed files</summary> ➕ `tinytorch/tests/08_training/test_training_coverage.py` (+480 -0) </details> ### 📄 Description ## Summary Six areas of Module 08 were fully implemented but had zero test coverage. This PR fills those gaps with a dedicated `test_training_coverage.py`. ## Coverage | Area | What's tested | |------|--------------| | `CosineSchedule` | Start/mid/end values, monotonic decrease, past-end edge case, single-epoch schedule | | `clip_grad_norm` | Clipping, no-clip, multi-param global norm, direction preserved, zero-grad safety, returns original norm | | `save/load_checkpoint` | File created, all keys present, epoch/step/history/weights restored, training resumes correctly, creates parent dirs | | `Trainer.evaluate` | Finite loss, float return types, eval mode set, weights unchanged, history recorded per call, classification accuracy (correct and wrong) | | Scheduler integration | LR recorded in history, `optimizer.lr` updated to match schedule, decreases over epochs | | Gradient clipping integration | Training completes without crash, weights still update, tight clip limits update magnitude | | Train/eval mode switching | `model.training` correct after `train_epoch`, `evaluate`, and train/eval/train cycle | ## Files | File | Change | |------|--------| | `tinytorch/tests/08_training/test_training_coverage.py` | New: 478 lines, 36 tests | --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

GiteaMirror added the pull-request label 2026-04-27 17:25:52 -05:00

GiteaMirror closed this issue

2026-04-27 17:25:54 -05:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: github-starred/cs249r_book#8128