Engineering
Dispatches from the ULMFiT trenches
Don’t just look at the benchmark figures when you evaluate your language model. A subjective judgement can also lead to interesting insights.
Why ULMFiT can still matter: training and inference times
In real-life applications computational efficiency of ML models is as important as evaluation metrics
Resurrecting monolingual recurrent language models with ULMFiT for Tensorflow
Recurrent language models are still alive and kicking. We release pretrained Polish and English ULMFiT models for Tensorflow Hub.
Front-end evolution: towards performance and consistency
Some say that working on software architecture is like gardening. Since there’s no such thing as “the final season” or “final weather”, all you can do is to check the forecast for the upcoming days and introduce changes through a long list of iterations.
Continual Learning
Continual learning is a machine learning domain that aims to mitigate catastrophic forgetting and enable models to be trained with an incoming stream of training data.
From research to production and back: Machine Learning with AWS
Let's see how Machine Learning can be built in the cloud. We’ll explore AWS services that will help you boost your productivity on the path from research to production.
DynamoDB - Cost reduction of scan() method
In our company, some time ago we started receiving alarms about exceeding the quota for one of the tables in Amazon DynamoDB. This caused alerts in other parts of the system as the table is used by many different processes.
Word2vec: Loss Function
The difference between probabilistic output vectors and target vectors is crucial for the network to learn. How can you measure this difference? What does the learning process look like? How do we indicate when our model is smart enough?
