DEDUPED Hot AI repos, turned into use cases Get 5 a day

How to run a smaller or quantized model on modest hardware with Rizzo Flow

Pick a lighter model size or quantization to run decisions on a laptop with limited memory.

Who it is forA person with a laptop or a CPU-only machine who wants to try typed decisions without the 4B default model.

The problemThe default 4B Q8_0 model is about 4.4 GB, which may be too large or too slow on a weak machine.

The outcomeA running server with a smaller model, and a measured comparison of its answers against the default on your own workload.

What you need first

  • Python 3.11 or newer
  • git
  • uv (https://docs.astral.sh/uv/)
  • About 4.4 GB of free disk space for the default 4B model, plus the llama.cpp runtime (about 11 MB on a Mac, 30 MB with Vulkan, about 570 MB with CUDA)
  • A machine that can run llama.cpp: an Apple, NVIDIA, AMD or Intel GPU, or no GPU at all

Steps

  1. Run uv sync --locked from the Rizzo Flow folder.
  2. Run uv run rizzo devices to see what the runtime detects and which device --device auto will use. (inferred, check the README)
  3. Run uv run rizzo download --size 1.7b for a faster, smaller first look.
  4. Run uv run rizzo serve --size 1.7b or uv run rizzo serve --quant q4_k_m to choose the model and quantization. (inferred, check the README)
  5. Download the Q4_K_M files first with uv run rizzo download if the quantization is not present yet, then run the same test questions on each option. (inferred, check the README)

Prompt for your agent

In the rizzo-flow repo, run `uv run rizzo devices`, then start the server with `uv run rizzo serve --size 1.7b` and send the same three questions from <path-to-test-file> that were answered by the default model. Return the answers and probabilities side by side. Verify that accuracy differences are recorded before choosing a model.

Caveat. The README says the 1.7B model is much less accurate and that quantization and hardware change probabilities, so compare results on your own data before you choose one.

Share: X LinkedIn Hacker News Email Copy link

Disclaimer

Deduped lists open-source repositories that belong to other people. We do not own, maintain, audit, or vouch for any of them, and a repo appearing here is not an endorsement or a safety check. The use cases, steps, and prompts are written by AI and may be incomplete, outdated, or just wrong. Where a use case is marked "Seen in the wild", we link to public posts where someone described something similar, but we have not verified those posts and the use case itself is still our AI's own write-up. Use all of this at your own risk. Only run code, commands, or agents on systems and accounts you are authorized to use, and read the code and every command before you run it, especially anything that touches credentials, files, or the network. Nothing here is professional, security, legal, or financial advice. All trademarks and project names belong to their respective owners, and their use here does not imply affiliation. Content is provided as is, without warranties of any kind. Found a mistake, a safety problem, or something you want removed? Email [email protected] and we will look at it promptly. Full disclaimer page