Misc. bug: missing chat-template support in llama-run

[engelmi]

Name and Version

$ llama-cli --version
version: 4460 (ba8a1f9)
built with cc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 for x86_64-linux-gnu

Operating systems

Linux

Which llama.cpp modules do you know to be affected?

Other (Please specify in the next section)

Command line

llama-run granite-code

Problem description & steps to reproduce

The llama-run CLI currently doesn't take the chat template of a model into account. Thus executing llama-run on a model requiring such a template will fail.

First Bad Commit

No response

Relevant log output

failed to apply the chat template

[engelmi]

One possible solution could be to extend llama-run so that it implicitly searches for a <model-name>.ollama.template and applies it via common_chat_apply_template - or directly downloads and saves it together with the model with the mentioned name pattern (if build with LLAMA_USE_CURL). In addition, another CLI option such as --chat-template could be added to not rely on an implicit naming convention (and template location).

Although it might take me a while, I'd be up for implementing this - if it makes sense, of course.