transforEmotion 0.1.7

Release Highlights

This version bring's several new featurs.

• Comprehensive Evaluation Suite (9d5532e, 72f5e8c)
  • Introduced evaluate_emotions() with accuracy, precision/recall, macro/micro F1, AUROC, ECE, and Krippendorff metrics.
• FindingEmo Dataset Integration (9887ba0, 9d5532e)
  • Automated download/caching via download_findingemo_data(), plus quality checks and annotation loaders.
• Vision Model Registry & Non‑CLIP Models (c107997, a1c1c94)
  • Added registry helpers (add_vision_model() ,list_vision_models(), removal utilities) for managing vision backends.
  • Registered BLIP and ALIGN presets and refreshed Python pipelines for GPU detection and reticulate setup.
• New LLM Support for RAG (fd53c5e, e743032, 2c7a4b5, 79f874b)
  • Expanded RAG stack with Gemma 3, Ministral 3B, Qwen 3, and TinyLLaMA options, plus retriever registry (vector + BM25).
  • Added JSON parsing/validation, evaluation utilities, and quieter logging across the RAG workflow.
• VAD & Scoring Enhancements (2586e2f)
  • Launched map_discrete_to_vad() and vad_scores() for weighted NRC VAD mappings across transformer, image, and video outputs.
  • Extended test coverage to ensure consistent discrete→VAD conversions.
• Docs & Infrastructure (a1c1c94, 2c7a4b5)