Structured Pruning Learns Compact and Accurate Models

Keyphrases

• Compact Model 100%
• Accurate Model 100%
• Distillation 100%
• Structured Pruning 100%
• Pruning Strategy 66%
• Unlabeled Data 66%
• Distillation Method 66%
• Latency 33%
• Knowledge Transfer 33%
• Parallelizable 33%
• Yield Model 33%
• Model Size 33%
• Neural Language Models 33%
• Hidden Unit 33%
• Different Granularity 33%
• Distillation Strategies 33%
• SQuAD Dataset 33%
• Model Training 33%
• Model Compression 33%
• Module Unit 33%
• Pruned Model 33%

Computer Science

• Unlabeled Data 100%
• Subnetwork 50%
• Granularity 50%
• Language Modeling 50%
• Knowledge Transfer 50%
• Model Compression 50%
• Pre-Trained Model 50%