paper review: “Graph Attention Networks”

arxiv: https://arxiv.org/abs/1710.10903 key points introduce “graph attention network(GAT)” which consists of “graph attention layers” which incorporate the “self-attention” idea from transformers to graph neural network the graph attention layers work by calculating weights of a node’s neighbors from the features of the neighbors by taking other neighbor’s existence into account. Read more…

Properly setting dataloader and callback for validation in pytorch DDP

pytorch distributed data parallel(DDP) is very useful and relatively well provided for creating a distributed training setup. However, the provided documentations and tutorial are mostly about “training” part and didn’t talk much about validation callbacks that run during training.

It is easy to think just using DistributedSampler for the validation dataloader would do all the work for you like it did in training dataloader, but it doesn’t. There are two main problems.

(more…)

paper review: “BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension”

arxiv: https://arxiv.org/abs/1910.13461 key points propose autoregressive model named BART, which is architecturally similar to standard transformer encoder + decoder Check out 5 pretraining tasks, and experiment which pretraining task is most helpful test BART performance with large scale pretraining on downstream tasks Model Architecture This work introduces BART, which is fundamentally Read more…

paper summary “Perceiver IO: A General Architecture for Structured Inputs & Outputs”

arxiv: https://arxiv.org/abs/2107.14795 Key points developing upon the Perceiver idea, Perceiver IO proposes a Perceiver like structure but where output size can be much larger and still keep overall complexity linear. (Checkout summary on Perceiver here) same with Perceiver, this work use latent array to save input information and run this through multiple self Read more…

paper summary: Swin Transformer: Hierarchical Vision Transformer using Shifted Windows

arxiv: https://arxiv.org/abs/2103.14030 Key points multi scale feature extraction. Could think of as adoption of FPN idea. restrict transformer operation to within each window and not the entire feature map → allows to keep overall complexity linear instead of quadratic apply shifted window to allow inter-window interaction fuse relative position information in Read more…