Convenience Functions for Common Models — convenience-models • wizrd

These functions provide convenient access to common local models.

llama(server = ollama_server(), temperature = 0, ...)
llamafile_llama(temperature = 0, ...)
llama_vision(server = ollama_server(), temperature = 0, ...)
nomic(server = ollama_server(), temperature = 0, ...)

Arguments

server: The Ollama-based models require an Ollama server to be running. By default, this calls ollama_server to start the server, if it is not already running. See the cautionary note in ollama_agent for the potential pitfalls of this convenience.
temperature: The temperature of the model, defaulting to zero for convenient stability during testing and demos.
...: Additional model parameters, see LanguageModelParams.

Details

llama, llama_vision and nomic are all pulled from Ollama, so Ollama must be installed.

llamafile_llama is meant as a quick start. It will download and cache a self-contained, cross-platform llamafile binary using the same weights as llama. No Ollama required.

Issues with llamafile

See llama_cpp_agent for resolving OS-specific issues with llamafile_llama.

Value

An Agent

Author

Michael Lawrence

Examples

if (FALSE) { # \dontrun{
    llamafile_llama() |>
        instruct("Answer questions about the mtcars dataset:", mtcars) |> 
        predict("Relationship between hp and fuel efficiency")
} # }