OpenAssistant provided a xor patch for the original LLaMa 30B model to produce a checkpoint of the model they are using on open-assistant.io.
This model feels like it performs better than the base LLaMa 30B model, so it would be great to have as an option to easily install and use via dalai, if possible.
OpenAssistant provided a xor patch for the original LLaMa 30B model to produce a checkpoint of the model they are using on open-assistant.io.
This model feels like it performs better than the base LLaMa 30B model, so it would be great to have as an option to easily install and use via dalai, if possible.