Tips

Photo Credit This is a short post describing how to use half precision in TorchScript. This can speed up models that were trained using mixed precision in PyTorch (using Apex Amps), and also some of the model trained using full precision (with some potential degradation of accuracy). TorchScript is a way to create serializable and optimizable models from PyTorch code. Any TorchScript program can be saved from a Python process and loaded in a process where there is no Python dependency. source ...

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 Motivation Apache Airflow is great for managing scheduled workflows, but in a lot of cases, it is an overkill and brings unnecessary complexity to the overall solution. Cron jobs are much easier to set up, have built-in support in most systems, and have a very flat learning curve. However, the lack of monitoring features and the consequential silent failures can be the bane of system admins’ lives. We want a simple solution that can help admins monitor the health of cron jobs in simple scenarios that do not warrant Airflow. The simple scenarios have the following characteristics: ...

This is a short post describing how to use Plotly to make text-heavy charts cleaner in R. Introduction David Robinson presented a beautiful way to visualize the ratings of the Office episodes in this screencast: The chart (shown below) is sufficiently readable when zoomed in on a full HD monitor, but is quite messy when exported to a smaller frame. Moreover, some of the episode names are not displayed (to avoid overlapping). ...

Photo Credit Motivation I was recently given a $300 credit to Google Cloud Platform from Google to participate in a Kaggle competition. This gives me free access to the powerful GPUs (T4, P100, even V100) to train models and thus opens the window to many new possibilities. However, the problem is that $300 can be used up rather quickly. For example, one Tesla P100 GPU cost $1.46 per hour, so $300 can only give me 200 hours or 8.5 days. Don’t forget there are still other costs from CPU, memory, and disk storage. ...