This is a guide to publishing a package to PyPI using Travis on tagged commits. Package version will automatically be decided by tagged commits so this will reduce your efforts to replace package version in setup.py The requirements are: A GitHub repository with a python package Travis command line tools …

A recent issue of the biomedical field is that the number of papers submitted is very large. As for PubMed, which is a database of medical literatures, the number is known as hundreds of thousands of documents are added each year. Because of the large number of articles submitted, it…

On May 1st, AWS announced new features of SageMaker’s Object2vec, which include automatic negative sampling, weight sharing, speeding up of training. The following Notebook is a tutorial for new features, it includes computing document embeddings, deploying trained models, and document retrieval. object2vec_document_embedding.ipynb We can understand how to use the features…

In named-entity recognition, f1 score is used to evaluate the performance of trained models, especially, the evaluation is per entity, not token. The function to evaluate f1 score is implemented in many machine learning frameworks. However, its target is classification tasks, not sequence labeling like named-entity recognition. Fortunately, Keras allows…

In machine learning, especially deep learning, GPU is often used to train machine learning models. In the most cases, GPU instances on the cloud services are selected by engineers because they are easier to prepare. But since the GPU instance is expensive, it is a waste of cost if you…

TL;DR I want to debug a model created with Keras. Keras doesn’t have a built-in debugger. Wrap Keras’ Session object with TensorFlow debugger(tfdbg). Keras, open-source neural networks library, is based on computation graphs. A typical Keras program consists of two separate stages: Build the neural network model with Sequential Model or…