LLM Learning Hub

workspace/llm-course/home

Dataset

A dataset is the collection of examples a model learns from, typically split into training, validation, and test sets. Quality, size, and diversity of the dataset determine model performance.

What is Dataset?

A dataset is the collection of examples a model learns from, typically split into training, validation, and test sets. Quality, size, and diversity of the dataset determine model performance.

A dataset is the collection of examples a model learns from, typically split into training, validation, and test sets. Quality, size, and diversity of the dataset determine model performance.

Where is it used?

GPT-3 was trained on Common Crawl, WebText2, books, and Wikipedia; Llama 3 used 15T tokens curated from web, code, and multilingual sources.

How to build it

Use `torch.utils.data.Dataset` and `DataLoader` to wrap a tensor or folder of files, set `batch_size` and `shuffle=True`, and iterate batches in a training loop.