Add runtime model compilation options

[mbsantiago]

Summary

• Add opt-in runtime compilation for training and inference via compile_model
• Compile the detector used by training and prediction
• Add BatDetect2API.compile() and compile_model=True options for API construction
• Allow passing a pre-built Lightning logger into the training workflow
• Add eta_min support to the cosine annealing scheduler config
• Document the new runtime compilation options

Testing

• just test
• just docs

Full test result: 522 passed

[Copilot]

Pull request overview

This PR adds opt-in runtime compilation support across training and inference, exposing compilation controls via configs and the v2 API, and documents the new options.

Changes:

• Introduces compile_model runtime option for both training (TrainingConfig) and inference (InferenceConfig), plus BatDetect2API.compile() and constructor flags to auto-compile.
• Adds float32 matmul precision control for training runs and supports providing a pre-built Lightning logger.
• Extends cosine annealing scheduler config with eta_min, updates tests, and documents the new runtime options.

Reviewed changes

Copilot reviewed 15 out of 15 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
tests/test_train/test_lightning.py	Refactors smoke config creation and adds tests for training-time compilation + matmul precision.
tests/test_inference/test_batch.py	Adds batch inference tests for compile-on-demand and no-recompile behavior.
tests/test_api_v2/test_api_v2.py	Adds API-level tests for explicit compilation and compile-on-construction.
tests/conftest.py	Adds a fixture to record detector compilation/call behavior; minor waveform generation refactor.
src/batdetect2/train/train.py	Adds optional Lightning logger injection, matmul precision setting, and optional detector compilation.
src/batdetect2/train/schedulers.py	Adds eta_min support to cosine annealing scheduler config/builder.
src/batdetect2/train/config.py	Adds runtime fields (compile_model, matmul precision) to TrainingConfig.
src/batdetect2/models/init.py	Adds compile_model() helper and exports it.
src/batdetect2/inference/config.py	Adds compile_model flag to InferenceConfig.
src/batdetect2/inference/batch.py	Compiles the detector prior to batch inference when requested.
src/batdetect2/api_v2.py	Adds BatDetect2API.compile(), compile-on-construction flags, and supports passing a pre-built train logger.
justfile	Adds -v to the example training invocation.
docs/source/reference/configs/training/training-config.md	Documents training runtime compilation and matmul precision options.
docs/source/reference/configs/inference/inference-config.md	Documents inference runtime compilation and API entry points.
docs/source/reference/api.md	Documents compile-on-construction and explicit compilation via API.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.