Does the model in this framework have to be trained to have conversations?

[code-isnot-cold]

I am new to using LLM, and is it possible to just download the open source LM and have multiple conversations without fine-tuning?

[Trace2333]

Of course you can. The finetuning aims to adjust the actor-model for better performance. If the DeepSpeed team opensource the model parameters, you can just download the model and use it without training.