Podcast Beta
Questions and Answers
Fine-tuning with custom data enables the model to better understand and respond to ______ jargon or regional language nuances.
industry-specific
Training a classifier with transformer-based models generally follows a ______-step approach.
two
During training, we will “freeze” the model such that its ______ will not be updated.
weights
A ______ Forward Neural Network is typically chosen as the classifier.
Signup and view all the answers
In this context, the dataset is the ______ dataset.
Signup and view all the answers
It transforms raw input data into a format that the model can understand, which may involve ______, padding, and batching.
Signup and view all the answers
Unlike traditional ______ methods which pad every sequence to the length of the longest one in the dataset, dynamic padding adjusts padding for each batch separately.
Signup and view all the answers
When freezing model ______, it’s generally better to freeze lower weights near the beginning of the model.
Signup and view all the answers
______ – A configuration object provided by the Hugging Face that holds our hyperparameters and options for the training process.
Signup and view all the answers
______ – A utility provided by the Hugging Face library that manages the fine-tuning process of a model.
Signup and view all the answers
