Today was for sure a coffee day vs. espresso. That happens sometimes, but it is not always apparent at the start. Starting off a coffee day with two shots of espresso is inherently problematic. A coffee day should start off a little bit slower and be a little more sustained. My big plan for this more sustained morning of sipping coffee will be to work out of the Google Colab environment on some GPT-2 efforts. Digging into the natural language processing space is where my focus is at today. Taking that course of action might mean that this weblog post is going to be a little bit on the shorter side, but it is entirely possible that the effort will help spark some inspired prose.
I have a working example setup that uses the OpenAI GPT-2 model. That is not novel or all that interesting. People have been doing that for months. I’m messing around with the code in the Colaboratory (Colab) environment. The benefit of that course of action is that the code is stored in a Jupyter notebook and is hosted by Colab. Working with other people in an environment that is very easy to replicate remains important to me going forward. Right now I’m working on digging into a method to take my writing corpus and store it in a basic text file online. That prevents people from having to scrape this weblog to get all the content. Working with the data from the export to clean it up for use as a GPT-2 model is taking a little bit of time, but it is a good learning experience. Getting that file setup and making it easily fetchable will allow me to update a version of a Jupyter notebook to train the GPT-2 model on that corpus.
That is what I’m working on at the moment.