mirror of https://github.com/azaion/ai-training.git synced 2026-04-22 11:46:35 +00:00

Files

T

Oleksandr Bezdieniezhnykh 222f552a10 [AZ-171] Add TensorRT tests, AC coverage gate in implement skill, optimize test infrastructure

- Add TensorRT export tests with graceful skip when no GPU available
- Add AC test coverage verification step (Step 8) to implement skill
- Add test coverage gap analysis to new-task skill
- Move exported_models fixture to conftest.py as session-scoped (shared across modules)
- Reorder tests: e2e training runs first so images/labels are available for all tests
- Consolidate teardown into single session-level cleanup in conftest.py
- Fix infrastructure tests to count files dynamically instead of hardcoded 20

Made-with: Cursor

2026-03-28 21:32:28 +02:00

955 B

Raw Blame History

Implementation Report — Dynamic Batch Export

Date: 2026-03-28 Epic: AZ-164 (Code Improvements) Total Tasks: 1 Total Batches: 1 Commit: 433e080

Summary

Enabled dynamic batch size for all three model export formats (ONNX, TensorRT, CoreML) by adding dynamic=True to the ultralytics export calls. TensorRT max batch size set to 8.

Tasks Implemented

Task	Name	Complexity	Status
AZ-171	dynamic_batch_export	2	Done

Changes

File	Change
src/exports.py	Added `dynamic=True` to export_onnx, export_tensorrt, export_coreml; changed TensorRT batch from 4 to 8
_docs/02_document/architecture.md	Updated Model Artifacts table to reflect dynamic batch support

Test Results

48 passed, 14 skipped, 6 errors (pre-existing ModuleNotFoundError for onnx package in e2e tests — environment dependency, not introduced by this change)

955 B Raw Blame History