Temperature in Text Generation

The speed at which the model generates tokens The maximum number of tokens the model can generate The randomness of the probability distribution over the next token The GPU thermal threshold during inference