Pytorch

Photo Credit Introduction Self-supervised learning made transfer learning possible in NLP [1] (by using language modeling as the pre-training task) and has started to show some potential in CV as well [2, 3, 4]. They make the downstream tasks more label efficient, that is, requires fewer labeled examples to achieve good prediction accuracies. In CV, we are already quite familiar with transfer learning from models pre-trained on the labeled Imagenet dataset. However, if the dataset used in the downstream task is significantly different from the Imagenet, transfer learning/fine-tuning usually would not be very helpful. ...

Photo Credit Overview Motivation Transformer models[1] have been taking over the NLP field since the advent of BERT[2]. However, the high numbers of parameters and the quadratically scaled self attention that is expensive both in computation and memory[3] make the modern transformer models barely fit into a single consumer-grade GPU. Efforts have been made to alleviate this problem[3][4][5], but they are still far from ideal: No public models that are pre-trained on BERT-scale corpus (at the time of writing). [3] The complexity of the public models is no smaller than the existing transformer models. [4] They are just smaller versions of BERT. The self attention is still quadratically scaled. [5] To make the inference possible on weaker machines, one of the more ideal solutions is to distill the knowledge of a fine-tuned transformer model into a much simpler model, e.g., an LSTM model. Is it possible? Tang et al.[6] shows that they can improve the BiLSTM baseline by distillation and some data augmentation. Although their accuracies are still lagging behind ones of the transformer models, it is still a promising direction. ...

Photo Credit Introduction AWS recently released TorchServe, an open-source model serving library for PyTorch. The production-readiness of Tensorflow has long been one of its competitive advantages. TorchServe is PyTorch community’s response to that. It is supposed to be the PyTorch counterpart of Tensorflow Serving. So far, it seems to have a very strong start. This post from the AWS Machine Learning Blog and the documentation of TorchServe should be more than enough to get you started. But for advanced usage, the documentation is a bit chaotic and the example code suggests sometimes conflicting ways to do things. ...

Photo Credit Synopsis I have the task of finding similar entries among 8,000+ pieces of news, using their title and edited short descriptions in Traditional Chinese. I tried LASER[1] first but later found Universal Sentence Encoder[2] seemed to work slightly better. Results from these unsupervised approaches are already acceptable, but still have occasional confusion and hiccups. Not entirely satisfied with the unsupervised approaches, I collected and annotated 2,000 pairs of news and fine-tuned the BERT model on this dataset. This supervised approach is visibly better than the unsupervised one. And it’s also quite sample-efficient. Three hundred and fifty training example is already enough to beat Universal Sentence Encoder by a large margin. ...

Photo Credit Synopsis Do you want multilingual sentence embeddings, but only have a training dataset in English? This post presents an experiment that fine-tuned a pre-trained multilingual BERT model(“BERT-Base, Multilingual Uncased” [1][2]) on monolingual(English) AllNLI dataset[4] to create sentence embeddings model(that maps a sentence to a fixed-size vector)[3]. The experiment shows that the fine-tuned multilingual BERT sentence embeddings have generally better performance (i.e. lower error rates) over baselines in a multilingual similarity search task (Tatoeba dataset[5]). However, the error rates are still significantly higher than the ones from specialized sentence embedding models trained with multilingual datasets[5]. ...