xiusu

Results 7 comments of xiusu

Thanks for your issue. To implement the training and search of supernet, we need to set the head number of each batch. Therefore, the first "head_dim" is only used to...

Thanks for your question. To implement the retraining process of a searched architecture, you can refer to config/retrain/ViTAS_1G_retrain.yaml. As in lines 82 and 122, the "net_id" defines the retrained architecture...

Thanks for your question. I leveraged 32 X V100 cards with 32G GPU RAM each to implement the search.

It takes about 2-3 days for searching an ViT architecture.

Yes, during sampling, the accuracy of ViT architecture is relatively low in supernet.

Thank you very much for your question. This code is the first version of ViTAS last year, and it is based only on BCNet and AutoSlim. This year I updated...

Thanks for your attention. Because the original code is too complicated, especially the Monte Carlo tree part, we're tidying up this code to make it more readable. We'll open-source the...