Commit History
-
🚀 SkyPilot Multi-Cloud GPU Support + Synthetic Data Generation
marcosremar +1: ClaudeImplemented complete infrastructure for training and annotation with multi-cloud spot instances. ## New Features ### 1. SkyPilot Integration (scripts/cloud/) - ✅ `skypilot_finetune.yaml` - Single GPU fine-tuning - ✅ `skypilot_multi_gpu.yaml` - Multi-GPU (8x) parallel training - ✅ `skypilot_annotate_orpheus.yaml` - Dataset annotation (118k samples) **Benefits**: - Automatic cheapest spot instance search across AWS/GCP/Azure - Up to 70% cost savings vs on-demand - Auto-recovery if preempted - Multi-GPU support (8x faster training) ### 2. Synthetic Audio Generation (scripts/data/) - ✅ `create_synthetic_test_data.py` - Generate emotion-like audio - 7 emotions: neutral, happy, sad, angry, fearful, disgusted, surprised - Configurable samples per emotion - Realistic acoustic characteristics: - Pitch modulation (vibrato/tremolo) - Harmonic structure - ADSR envelopes - Emotion-specific features **Usage**: ```bash python scripts/data/create_synthetic_test_data.py --samples 50 ``` #…13e402e -
Implement OPTION A: Optimized 3-model ensemble with fine-tuning
marcosremar +1: Claude## Changes ### Core Ensemble Optimization - Updated Balanced mode to use 3 diverse models (optimal per research) - Reweighted models: emotion2vec (0.50), Whisper (0.30), SenseVoice (0.20) - Emotion2VecModel now auto-loads fine-tuned weights if available ### Fine-tuning Infrastructure - Added finetune_emotion2vec.py: Complete fine-tuning script - Supports VERBO (1,167) + emoUERJ (377) datasets - Data augmentation: time stretch, pitch shift, noise injection - Automatic train/validation split - Per-class accuracy metrics - Added download_ptbr_datasets.py: Dataset downloader/helper - Instructions for VERBO, emoUERJ, CORAA-SER - Local dataset preparation from folders - HuggingFace dataset creation ### Documentation Updates - README: Added scientific rationale for OPTION A - 10+ academic papers (2024-2025) justification - Cost-benefit analysis: 3 vs 5 models - Updated model weights and configuration - QUICKSTART: Added fine-tuning workflow - PROJECT_SUMMARY: Comprehens…edb9bc2