Found 4 related articles
Temperature and Top-p both control the model's "how to choose the next word". To put it simply, they affect how conservative and divergent the answers...
System Prompt can be understood as "rules placed at a higher level". It is not the prompt phrase that ordinary users enter casually, but the basic ins...
Tokens can be understood as the "smallest unit of work" when the model processes text. It doesn't necessarily equal a word, a word, or a punctuation, ...
The attention mechanism simply means that the model learns to "look at the key points" when processing information. When a person reads a paragraph, t...