Hugging Face Made Crystal Clear · chapter 3: Zero to First Inference
The box, opened
2026-10-31
A tokenizer converts your text to numbers, the model predicts more numbers, and post-processing turns them back into language.
Below: the paragraphs from the book that build this idea, then the diagram itself (Figure 3.2), and a recap. About a minute of reading.
Run it, and watch what happens the first time, because the first time is special. Progress bars appear: the config, the tokenizer files, and one model.safetensors file of about 724 MB come down from the Hub. Then, after a pause while the weights load, your prompt comes back with new words attached to it. Words that were computed on your machine, by a model you now own a copy of.
What did pipeline() actually do? Three jobs that every tutorial that skips it makes you do by hand. Figure 3.2 opens the box.
Recap
- The idea: A tokenizer converts your text to numbers, the model predicts more numbers, and post-processing turns them back into language.
- The picture: Figure 3.2, from chapter 3 ("Zero to First Inference") of Hugging Face Made Crystal Clear.
- Go deeper: the chapter builds this step by step, with recipes and sources at the end.
This diagram is one of many in Hugging Face Made Crystal Clear.
Every chapter opens with the gist, draws the hard ideas, and ends with recipes and sources.
Get the book
