Code4Talks
More details in OICampus channel
Distrito Telefónica. Innovation & Talent Hub
In this episode of #Code4Talks, Fran Ramírez and Pablo Gómez Álvarez explore how to develop and deploy artificial intelligence applications. Pablo, a student of Data Science and Artificial Intelligence at the Universidad Politécnica de Madrid, shares his experience as a developer of backend services and machine learning models.
The episode focuses on the implementation of a complete pipeline using the Hugging Face library. Pablo explains step by step how to program an artificial intelligence application, focusing on natural language processing and sentiment classification. To train the model, a dataset called "emotions" from Hugging Face is used. Special emphasis is put on the tokenisation and inference phases, which are fundamental in the pipeline.
In addition, Pablo introduces the concept of "fine-tuning" by choosing a pre-trained model, specifically, DistilBERT. This allows the training process to be accelerated. The hyperparameters used, such as the number of epochs and batch size, are detailed, and the importance of checkpointing to save time and resources in training is also explained.
Finally, Pablo shows how to convert the trained model into a pipeline accessible to non-programmers. To do this, the "gradio" library is used to create a graphical interface that allows users to interact with the model and predict the emotion associated with a text without the need for programming knowledge.
In summary, the episode deals with the practical implementation of an artificial intelligence pipeline, from the choice of the model to the creation of a user-friendly interface for end users. It offers a complete overview of the process, from the point of view of an experienced developer.
More details in OICampus channel
