genta 0.0.3.9

Simple python interface to access models from Genta API.

genta-api

Genta Python API Package.

Installation

This API wrapper can be installed by cloning the repository and running pip install . in the root directory. You can also use pip install genta

Examples

You can use this module like so:

from genta_api import GentaAPI
api = GentaAPI(token='your-token')
response, statusCode = api.ChatCompletion(
    chat_history=[
        { "role": "user", "text": "Hello, how are you?" },
        { "role": "assistant", "text": "I'm fine, thank you." },
        { "role": "user", "text": "That's good to hear." },
    ],
    model_name='Starstreak',
    best_of=1,
    decoder_input_details=True,
    details=True,
    do_sample=True,
    max_new_tokens=128,
    repetition_penalty=1.03,
    return_full_text=False,
    seed=None,
    stop=[],
    temperature=0.7,
    top_k=50,
    top_n_tokens=None,
    top_p=0.95,
    truncate=None,
    typical_p=0.95,
    watermark=False,
)
print(response, statusCode)

A variety of options can be configured when calling the ChatCompletion method:

• chat_history: a list of dictionaries, each containing a "role" (user, assistant, or system) and "text".
• model_name: the name of the machine learning model to use. By default, it uses 'Starstreak'.
• best_of: an integer specifying the number of responses the model should generate. The best one will be chosen. Default is 1.
• decoder_input_details, details, do_sample: boolean flags to control various model behaviors.
• max_new_tokens: the maximum output length that the model will produce.
• repetition_penalty: a penalty for repetitive output. Default is 1.03.
• return_full_text: whether the output includes original input text as well. Default is False.
• seed: an integer to seed random responses.
• stop: a list of stop tokens.
• temperature: a measure of randomness in the output.
• top_k, top_n_tokens, top_p, truncate, typical_p: control parameters for the output.
• watermark: whether to include a watermark in the output. Default is False.