DIY LLMs with Modal

Running your own LLMs is harder than making an API call, but sometimes hard things are worth doing. Tips and tricks for self-hosting LLM inference using Modal, where provisioning GPUs is just an API call.