Hence, these redundant channels are pruned, which may result in a slight loss in accuracy, but this loss can potentially be compensated for during the subsequent fine-tuning process. After sparse training, a pruning threshold is determined based on the specified pruning rate, and BN layer ...
Latest commit chrislusf reduce dependency Nov 23, 2020 73d4234·Nov 23, 2020 History History
Hence, these redundant channels are pruned, which may result in a slight loss in accuracy, but this loss can potentially be compensated for during the subsequent fine-tuning process. After sparse training, a pruning threshold is determined based on the specified pruning rate, and BN layer ...