This is Kakugo 3B Lao, a small language model (SLM) fine-tuned to interact with the user in Lao. For Kakugo in other languages, check out the model and dataset collections. To use this model, you can use your preferred LLM inference package. This model should work with any package that supports the original base model ibm-granite/granite-4.0-micro. We provide examples for how to run this with Huggingface or vLLM: Then run the following Python code to generate a response from the LLM. N.B. — We recommend using a repetitionpenalty` of 1.05 as sometimes the model can stuck in a loop of generating repetitive text when generating low-resource languages. You can set doreasoning` to be either True or False to turn «thinking mode» on or off, respectively. If the model is used in thinking mode, then it will take longer to generate a response, but may lead to a better generated response. This mode is still experimental, so try both using and not using it for your use-case. Then run the following Python code to generate a response from the LLM. N.B. — When using vllm for inference of multiple inputs, we recommend inputting them all at once. I.e., add more items to the outer list of the…
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: ptrdvn
Теги: granitemoehybrid, low-resource-language, data-distillation, conversation, lao, Lao, conversational
Лайков: 3 | Загрузок: 92
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
