Toan-Minh-Duong-Son/ast-audio-analysis-multilabel-model

AST Multi-label Audio Classification Model cho EchoVerse

Thông tin mô hình

• Tên: ast-multilabel-model
• Base model: MIT/ast-finetuned-audioset-14-14-0.443
• Ngày huấn luyện: 2025-07-17 09:38:29
• Task: Multi-label audio classification
• Framework: PyTorch + Transformers

Cấu hình

• Sample Rate: 16000Hz
• Max Audio Length: 10.0s
• Learning Rate: 5e-05
• Weight Decay: 0.0001
• Epochs: 5
• Batch Size: 16
• Data Fraction: 0.05

Nhãn phân loại

• Genre Tags (16): Pop, Rock, Jazz, Classical, EDM, Hip-hop, R&B, Country, Folk, Ballad, Dance, Chillout, Ambient, Funk, Soul, Instrumental
• Mood Tags (16): Excited, Energetic, Happy, Joyful, Angry, Frustrated, Tense, Aggressive, Sad, Depressed, Melancholic, Bored, Relaxed, Peaceful, Calm, Content
• Instrument Tags (17): Guitar, Piano, Violin, Bass, Harp, Flute, Saxophone, Trumpet, Clarinet, Drums, Percussion, Timpani, Cymbals, Bells, Xylophone, Synthesizer, Electric Guitar
• Vocal Tags (12): Male, Female, Duet, Chorus, Instrumental, Rap, Acapella, High-pitched, Low-pitched, Falsetto, Vibrato, Whisper

Hiệu suất mô hình

genre

• f1_micro: 0.0000
• f1_macro: 0.0000
• precision: 0.0000
• recall: 0.0000
• hamming_loss: 0.0000

mood

• f1_micro: 0.0000
• f1_macro: 0.0000
• precision: 0.0000
• recall: 0.0000
• hamming_loss: 0.0000

instrument

• f1_micro: 0.0000
• f1_macro: 0.0000
• precision: 0.0000
• recall: 0.0000
• hamming_loss: 0.0000

vocal

• f1_micro: 0.0000
• f1_macro: 0.0000
• precision: 0.0000
• recall: 0.0000
• hamming_loss: 0.0000