Use two running examples. ClothingCNN reads a clothing image and chooses a label. A chat language model reads token context and writes more text. Both are trained models. Their day-to-day work looks different.
Input typePixel grid
One stepLook at the image
Output typeClothing label
Input typeToken sequence
One stepChoose next token
Output typeMore text
The practical differences
- Input. ClothingCNN expects a prepared image grid. A chat model expects a sequence of tokens and often a large recent context.
- Output. ClothingCNN returns scores that the app maps to a label. A chat model returns a stream of text tokens.
- One step. ClothingCNN can classify one image in one model call. A chat reply usually grows through repeated next-token calls.
- Visible mistakes. ClothingCNN may call a shirt a coat when the image is blurry. A chat model may state a wrong fact in smooth, confident writing.
- What ships. A clothing app ships image capture, preprocessing, the model asset, and a label view. A chat app ships context handling, streaming text, limits, and a model service.
One model can take more than one kind of input
A multimodal system can accept an image and text together. It may answer a question about a photo. The useful habit stays the same. Name each input and output before you name the model.
The forward clothing course makes the image path concrete from Fashion-MNIST through SwiftUI. From the metal to the model shows how to look at the lower layers. Different architectures still need the same care when you profile them. Find out what the chip is doing.
Compare jobs, not labels. Ask what enters, what one step does, what leaves, and where a mistake appears.
Next, choose a model shape from the app feature you need.
Key concepts
- ClothingCNN: pixel grid in, one look, one clothing label out.
- A chat model: token sequence in, next token out, repeated until the reply is done.
- Compare input, output, step count, visible mistakes, and what the app ships.
- Classifier mistakes look like a wrong label. Chat mistakes can look like smooth wrong facts.
Takeaways
- You can compare a classifier and a chat model without using brand names.
- You can list what a clothing app ships versus what a chat app ships.
- You can ask what enters, what one step does, what leaves, and where mistakes show.