Text Generator Brain Text Generator


Best will choose the best model we have for the given input Instruct is good at following complex instructions Multilingual is better at code and other spoken languages than English Chat is good at role play, following creative instructions, can have spelling mistakes Fastest will take any model if its the fastest, this can cause inconsistency.

How many results should be generated

The maximum amount of tokens to generate. tokens are pieces of text around 4 letters each.

Stops generation once it has generated a number of sentences. stop_reason: max_sentences will be returned if generation stops because of the max sentences setting

The minimum probability of a token to be generated. 0 turns this off, small values let generation go longer. large values stop sooner when unlikely text is generated stop_reason: min_probability will be returned if generation stops because of the min probability setting

A list of sequences that will stop generation. stop_reason: SEQUENCE will be returned if generation stops because the given SEQUENCE was generated this sequence is not included in the generated results

Probability threshold for token selection, higher means higher diversity results

Number of top-probability tokens to keep when deciding which token to use higher means higher diversity.

Higher means more diversity smoothing the probability distribution of choices made Smaller warps the distribution to be sharper and generates more likely/consistent results

Stops the network repeating previously seen tokens. Higher values cause more creative writing, taking the conversation in new directions. Too high values may cause rambling, hard to understand text and forgetting previous important details.

A seed for the random number generator. This makes the results reproducible. 0 will not make results reproducible but will be slightly faster