Skip to main content

Pick a model

This tutorial takes one model from the catalog to a served endpoint in four parts, each a complete session, and ends with the same model running inside a C++ program. Core concepts are explained where they first appear. This part picks the model and learns everything about it without downloading a single weight.

It assumes the install directory exists and clika-modelverse is on your PATH (see Quick install). The model is Llama 3.2 1B Instruct, small enough to run on any machine in the system requirements; every command works the same with any other model in the catalog.

The catalog knows the families​

list prints every registered model family: its modalities, the commands it provides, and whether it is runnable on this build. No network is involved; the catalog is compiled into the clika-modelverse executable.

clika-modelverse list
registered model families (54 runnable; 9 hidden, --all shows all)

chatterbox  runnable
  *  Input Modalities: text
  * Output Modalities: audio
  *      Valid Combos: text -> audio
  *          Commands: speak, serve, bench
  *            Web UI: generic page
  *            Vendor: Resemble AI
-----------------------------------------
...

A family is the model architecture Modelverse knows how to run; a model you fetch is a checkpoint of that family. The Commands row is the contract for parts 2 and 3: whatever it lists is what that model can do.

A source names a model​

Everything model-specific starts from a source: a Hugging Face Hub repo id (<org>/<repo>), a pasted Hugging Face URL, or a local directory. Modelverse runs checkpoints in the formats model publishers ship on the Hub: safetensors, torch containers (pytorch_model.bin), and GGUF. (ONNX exports are a different lane: the ClikaRT runtime runs ONNX models directly; its how-to guide covers that path.) info resolves a source's identity:

clika-modelverse info meta-llama/Llama-3.2-1B-Instruct
meta-llama/Llama-3.2-1B-Instruct
  family=llama  model_type=llama  architecture=LlamaForCausalLM
  variant: dense
  components: tokenizer=yes image=no audio=no video=no
  companions:
    config.json                                           877 B
    generation_config.json                                189 B
    tokenizer.json                                      8.7 MiB
    tokenizer_config.json                              53.2 KiB
  weights:
    model.safetensors                                   2.3 GiB
  total: 2.3 GiB  (weights 2.3 GiB, 5 files)
  repository: 4.6 GiB in 13 files; the rest is not fetched (other weight formats, files the fetch never takes)
  (dry; nothing downloaded)

Identity resolution reads configuration files only. clika-modelverse downloads a few KB of JSON, matches it against the registered families, and reports what it found; weights do not move. This is deliberate: you can interrogate a 70B model from a laptop.

--dry shows what a fetch would cost​

Add --dry and info also lists the files a fetch downloads, with sizes, split into companions (configs, tokenizer) and weights, and says how much of the repository it leaves behind (other weight formats, files the fetch never takes):

clika-modelverse info meta-llama/Llama-3.2-1B-Instruct --dry
meta-llama/Llama-3.2-1B-Instruct
  family=llama  model_type=llama  architecture=LlamaForCausalLM
  variant: dense
  components: tokenizer=yes image=no audio=no video=no
  companions:
    .gitattributes                                      1.5 KiB
    LICENSE.txt                                         7.5 KiB
    README.md                                          40.8 KiB
    USE_POLICY.md                                       5.9 KiB
    config.json                                           877 B
    generation_config.json                                189 B
    original/params.json                                  220 B
    original/tokenizer.model                            2.1 MiB
    special_tokens_map.json                               296 B
    tokenizer.json                                      8.7 MiB
    tokenizer_config.json                              53.2 KiB
  weights:
    model.safetensors                                   2.3 GiB
    original/consolidated.00.pth                        2.3 GiB
  total: 4.6 GiB  (weights 4.6 GiB, 13 files)
  (dry; nothing downloaded)

Nothing here is Llama-specific. info google/gemma-3-4b-it reports variant: dense + multimodal; info openai/whisper-large-v3-turbo reports an audio component. Some repos ship several weight options to choose between; Run a specific GGUF quantization covers picking one.

The model's own commands​

clika-modelverse has four commands that work without naming a model (list, devices, info and fetch; generate-template, their file-form helper, rides beside them). Every other command runs on a model you name: pass the source and --help, and clika-modelverse resolves its family and prints that family's own command surface:

clika-modelverse meta-llama/Llama-3.2-1B-Instruct --help
meta-llama/Llama-3.2-1B-Instruct
  family=llama  text->text

usage: clika-modelverse meta-llama/Llama-3.2-1B-Instruct [options] <command>

commands the 'llama' family provides for this model

options:
  -h, --help     show this help and exit
  --verbose, -v  debug diagnostics (the effective-options report)
  --quiet, -q    warnings and errors only
  --no-color     plain output (also honored: $NO_COLOR, TERM=dumb, non-tty)

commands:
  prompt         one templated generation of the positional user message
  chat           interactive terminal chat (FTXUI ui; plain repl off a tty)
  serve          run the openai-compatible http server
  bench          the family-owned benchmark flow
  mm_bench       the multi-modality benchmark flow (media axes over the serving
                 path)

Asking a model for a command its family does not provide refuses precisely and names what it does provide, with exit code 3 (Script clika-modelverse has the full exit contract).

Next: part 2, the weights arrive and the model speaks.