Fixing corrupt rows

[Josef-Haupt]

No description provided.

[Copilot]

Pull request overview

This PR addresses data corruption and inconsistent duration calculations in the acoustic inference pipeline by making duration derivation deterministic (frames / samplerate) and by fixing tensor growth to preserve masking for unwritten segments.

Changes:

• Compute audio duration from frames / samplerate instead of relying on sf_info.duration.
• Replace in-place ndarray.resize() growth with explicit reallocation + copy to ensure newly added segments remain masked/uninitialized as intended.
• Add focused regression tests for duration computation and tensor resize/masking behavior (including the “initial pointer = 0” case).

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
src/birdnet/acoustic/inference/core/producer.py	Derives duration from sample frames and samplerate to avoid inaccurate sf_info.duration.
src/birdnet/acoustic/inference/core/prediction/prediction_tensor.py	Reworks segment capacity growth to keep unwritten prediction segments masked; adjusts initial allocation for zero pointer.
src/birdnet/acoustic/inference/core/encoding/encoding_tensor.py	Mirrors the prediction tensor resize fix for embeddings; adjusts initial allocation for zero pointer.
src/birdnet_tests/acoustic_models/inference/producer_py/test_get_file_segments_with_overlap.py	Adds regression test ensuring duration uses frames/samplerate rather than sf_info.duration.
src/birdnet_tests/acoustic_models/inference/predictions/prediction_tensor_py/test_prediction_tensor.py	Adds tests verifying resize preserves masking and that zero-pointer initialization starts empty.
src/birdnet_tests/acoustic_models/inference/encoding/encoding_tensor_py/test_encoding_tensor.py	Adds tests verifying resize preserves masking and that zero-pointer initialization starts empty.

---

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