Inference
Inference is when we use a trained model to produce an answer, a prediction, or any other output. The learning is already done, now the model simply puts what it learned to work.
Inference Playground
Turn a prompt into an output
Pick a prompt, then press Run inference. The same trained model reads your prompt and produces a response. Notice it gives different outputs for different prompts.
Choose a prompt
User Prompt
Trained Model
Ready, waiting for a prompt.
Output
Press Run inference to see the response.
The model is not learning here. It is using what it already learned.
Training vs Inference
Two very different jobs
Training is how a model learns. Inference is how we use the model it became.
Training
The model learns from data.
InputTraining examples
OutputAn improved model
Inference
The trained model is used.
InputA user prompt
OutputAn answer or prediction
What Happens During Inference?
From prompt to response, step by step
Every time you ask a chatbot something, these four steps quietly happen behind the scenes.
1
User gives input.
2
The trained model processes it.
3
The model predicts the output.
4
The user receives the response.
🔑
Key takeaways
- • Inference means using a trained model.
- • Training changes the model.
- • Inference uses the model to produce outputs.
- • Chatbots use inference when answering users.
- • Inference happens after training.
Summary
Training is how a model learns. Inference is how we use the trained model to get answers.
🎉 Training Fundamentals section complete