best_distilbert_model

Model Description

Intended Uses & Limitations

✅ Intended Use

• Primary Task: Sentiment analysis for album reviews.
• Dataset: Fine-tuned on 19,305 album reviews (binary labels: 1 = Positive, 0 = Negative).
• Ideal for: Music review sentiment analysis.

⚠️ Limitations

• May not generalize well to non-music-related reviews.
• Optimized for binary sentiment classification, not multi-class sentiment.

Training & Evaluation Data

Dataset Details

• Dataset Source: Pitchfork Album Reviews
• Training Set Size: 19,305 reviews
• Test Set Size: 1,566 reviews
• Labels: Binary classification (0 = Negative, 1 = Positive)

Evaluation Metrics

• Best Test Accuracy: 73.44%
• Best Generalization Settings:
  • Dropout: 0.2
  • Learning Rate: 5e-5
  • Batch Size: 16
  • Warmup Steps: 500

Training Procedure

Hyperparameters Used

• Learning Rate: 5e-5
• Train Batch Size: 16
• Eval Batch Size: 16
• Epochs: 2
• Weight Decay: 0.01
• Dropout: 0.2
• Optimizer: AdamW (betas=(0.9, 0.999), epsilon=1e-08)
• LR Scheduler: Linear
• Warmup Steps: 500

Framework Versions

• Transformers: 4.48.3
• PyTorch: 2.6.0+cu124
• Datasets: 3.4.1
• Tokenizers: 0.21.1

Performance Metrics

• Best Test Accuracy: 73.44%
• Evaluation Metrics Used: Accuracy
• Generalization Settings:
  • Dropout: 0.2
  • Learning Rate: 5e-5
  • Batch Size: 16
  • Warmup Steps: 500