T5

Model name: t5_local

About T5

T5 is a text-to-text transformer model that converts input text into output text. It is based on the Transformer architecture where encoder processes the input and a decoder then genrates the output. It is pre-trained on a large corpus of text data and is capable of performing a wide range of natural language processing tasks, such as chatbots, translation, summarization, and question answering.

Read more about T5 on Wikipedia.

Supported aidb operations

• encode_text
• decode_text
• encode_text_batch
• decode_text_batch

Supported models

• t5-small (default)
• t5-base
• t5-large
• t5-3b
• t5-11b

Registering the default model

SELECT aidb.register_model('my_t5_model', 't5_local');

Registering a specific model

SELECT aidb.register_model(
  'another_t5_model',
  't5_local',
  '{"model": "t5-large", "revision": "main"}'::JSONB 
)

Model configuration settings

The following configuration settings are available for T5 models:

• model - The T5 model to use. The default is t5-small.
• revision - The revision of the model to use. The default is refs/pr/15. TODO
• temperature - The temperature to use for sampling. The default is 0.0.
• seed - The random seed to use for sampling. The default is 1599222198345926291.
• max_tokens - The maximum number of tokens to generate. The default is 16384.
• repeat_penalty - The repetition penalty to use. The default is 1.1.
• repeat_last_n - The number of tokens to consider for the repetition penalty. The default is 64.

Model credentials

No credentials are required for the T5 model.