One could fine-tune LLMs for so-called tool use. See the docs regarding how to do this with TRL: Learning Tools (Experimental 🧪). It teaches an LLM how to use a calculator.
Note: this is pretty experimental.
One could fine-tune LLMs for so-called tool use. See the docs regarding how to do this with TRL: Learning Tools (Experimental 🧪). It teaches an LLM how to use a calculator.
Note: this is pretty experimental.