argmax — position of the largest
Read "arg max". It gives the position (index) of the largest value, not the largest value itself.
is the biggest logit. is which token has it. Greedy decoding outputs the token at every step.
The difference shows up in the safe softmax too: there you subtract , the largest value, from every logit to prevent overflow.
Example
Counting positions from 1. The largest value is 3.2; it sits at position 2.
Where the course uses it
Related terms
- lim — limitRead "the limit as T goes to 0". It is the value a quantity gets closer and closer to as something is pushed toward an extreme, even if that extreme can't be plugged in directly.
- TemperatureRead "z i over T". Each logit is divided by the temperature T before softmax. T below 1 makes the distribution sharper; T above 1 makes it flatter.
- SoftmaxRead "softmax of z". It turns any list of numbers into probabilities: every output is positive and they add up to 1. Bigger inputs get bigger shares.