Izvajalec / Course provider: Jožef Stefan Institute (JSI), Network Infrastructure Centre (CMI)
Predavatelji / Instructors: Špela Vintar (CMI JSI), Mojca Brglez (CMI JSI)
Learning objectives: Learning about the use of large language models via application programming interfaces (APIs), to gain an understanding that enables effective planning of automation of language and other tasks
Course content: Large language models can perform a whole range of language tasks, such as translation, revision, checking of professional terminology, summarizing or scraping data from text strings. It is often easiest to include a language model in the workflow via an API, which means that a specific language task that we want to perform repeatedly or on a larger amount of data can be automated with an appropriate call and a program interface to the selected non-local model.
Participants are first introduced to the technical aspects of accessing models via APIs, with practical examples of program scripts, and also by testing the design of calls for different input data formats and types of tasks. This is followed by a more in-depth practical part, where we present three examples of language tasks, compare different calls, and the results of the models accessed via the API interface. Participants are encouraged to prepare their own data before the training, on which they can test the presented methods.
In the last part of the training, we talk about evaluating the results, which usually includes creating an evaluation set of correct solutions (gold standard) and using various evaluation metrics to assess the performance of the model.
Learning outcomes: Understanding the technical aspects of accessing language models through programming interfaces, the ability to plan their own usage scenarios, prepare data, and evaluate results