Instructions to use poolside/Laguna-S-2.1-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use poolside/Laguna-S-2.1-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf poolside/Laguna-S-2.1-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf poolside/Laguna-S-2.1-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf poolside/Laguna-S-2.1-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf poolside/Laguna-S-2.1-GGUF:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf poolside/Laguna-S-2.1-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf poolside/Laguna-S-2.1-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf poolside/Laguna-S-2.1-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf poolside/Laguna-S-2.1-GGUF:Q4_K_M

Use Docker

docker model run hf.co/poolside/Laguna-S-2.1-GGUF:Q4_K_M

LM Studio
Jan
Ollama
How to use poolside/Laguna-S-2.1-GGUF with Ollama:
```
ollama run hf.co/poolside/Laguna-S-2.1-GGUF:Q4_K_M
```

Unsloth Studio

How to use poolside/Laguna-S-2.1-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for poolside/Laguna-S-2.1-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for poolside/Laguna-S-2.1-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for poolside/Laguna-S-2.1-GGUF to start chatting

How to use poolside/Laguna-S-2.1-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf poolside/Laguna-S-2.1-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "poolside/Laguna-S-2.1-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use poolside/Laguna-S-2.1-GGUF with Docker Model Runner:
```
docker model run hf.co/poolside/Laguna-S-2.1-GGUF:Q4_K_M
```

Lemonade

How to use poolside/Laguna-S-2.1-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull poolside/Laguna-S-2.1-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Laguna-S-2.1-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use poolside/Laguna-S-2.1-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf poolside/Laguna-S-2.1-GGUF:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default poolside/Laguna-S-2.1-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use poolside/Laguna-S-2.1-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf poolside/Laguna-S-2.1-GGUF:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "poolside/Laguna-S-2.1-GGUF:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

The size jump from previous q4_k_m GGUF is making this unusable on M5 Max 128GB

#18

by alxarch - opened 23 days ago

Discussion

alxarch

23 days ago

The last commit mentions:
gguf: replace Q4_K_M with mixed-precision Q4_K_Mx40-BF16x8 (bf16 sensitive layers)

That caused a huge size jump to 96GB for the 4bit variant (presumably because of the un-quantized layers).
The 96GB size is prohibitive for large context sizes on 128GB unified RAM.
Is this a mistake? Why not use a q8_0 or similar for the sensitive layers? Was just ready to switch from the unsloth quantization to the 'official' one but now I am not sure what to do. If someone from poolside could clarify the situation about what was the reason for switching sensitive layers to uncompressed, it would help us make informed decision about whether to use this or the unsloth UD-q4_k_xl to be able to take advantage of 256k context

sunny-lee-zen

23 days ago

Agreed, the jump from 68.2GB to 96GB is a bit extreme.

I can download the previous version from the git history, but I'm curious why the decision was made to bf16 sensitive layers from q8_0. Was the previous version not performing well enough?

jettoblack

22 days ago

The previous version fit perfectly on a 96GB VRAM or 128GB unified memory setup with 256k of FP8 context. That's a really popular class of setups, ranging from quad 3090/4090s, Blackwell 6000, plus all of the 128GB devices like Macbook Pro / Mac Studio, DGX Spark, AMD Strix Halo...

The new version doesn't fit on ANY of those... 😢

Yonisz

21 days ago

•

edited 21 days ago

I'm trying to fine tune an LLM that will detect that loops in real time and end reasoning when it's looping so it will generate an answer, i will update later.

IHaveNoClueAndIMustPost

18 days ago

I uploaded a version here with the BF16 parts quanted down to Q8 without touching anything else. Made the model fit again in 96GB VRAM and the higher precision should still make it an improvement over the straight Q4_K_M variant.

Yonisz

18 days ago

I uploaded a version here with the BF16 parts quanted down to Q8 without touching anything else. Made the model fit again in 96GB VRAM and the higher precision should still make it an improvement over the straight Q4_K_M variant.

I tested it, it still lopping

IHaveNoClueAndIMustPost

18 days ago

Since my quantization is nearly byte-identical to the official GGUF (only the later BF16 parts were requanted), it wouldn't change behavior present in the base model to any major degree.
I haven't had any looping issues myself but I'm running fairly conservative settings in a custom harness. I did however have issues getting the reasoning to work reliably as it performs quite poorly without it.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment