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

Libraries

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

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="poolside/Laguna-S-2.1-GGUF",
	filename="laguna-s-2.1-DFlash-BF16.gguf",
)

llm.create_chat_completion(
	messages = "No input example has been defined for this model task."
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

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

Hermes Agent new

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 new

OpenClaw new

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"

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

DFlash GGUF missing sliding-window metadata -> draft acceptance collapses beyond short contexts

#10

by Dreadbyte - opened about 8 hours ago

Discussion

Dreadbyte

about 8 hours ago

`laguna-s-2.1-DFlash-BF16.gguf` is missing sliding-window metadata → draft acceptance collapses to ~0% beyond short contexts

Summary

The published DFlash draft GGUF (laguna-s-2.1-DFlash-BF16.gguf) carries no
dflash.attention.sliding_window key. The drafter's HF config
(poolside/Laguna-S-2.1-DFlash/config.json)
declares all 6 layers as sliding_attention with window 512:

"layer_types": ["sliding_attention", ...x6],
"sliding_windows": [512, 512, 512, 512, 512, 512]

but the GGUF metadata contains only:

dflash.context_length = 1048576
dflash.rope.freq_base = 500000.0
(no dflash.attention.sliding_window, no pattern)

The laguna branch of poolsideai/llama.cpp treats SWA as optional when loading
the dflash arch (src/models/dflash.cpp, load_arch_hparams), so the drafter
silently runs full global attention. That is in-distribution for prompts
shorter than the 512-token window and increasingly out-of-distribution beyond
it.

Observed impact

llama.cpp (fork @ 04b2b72), Laguna-S-2.1 UD-Q3_K_M target, greedy sampling,
--spec-draft-n-max 15:

Context depth	mean accepted draft len (as published)	(after fix)
~70 tok	3.61	3.61 (unchanged, inside window)
~18k tok	1.01 (acceptance 0.07% — drafts never survive)	2.12

After the fix, a depth ladder over the same document (1.7k → 18k tokens) is flat
and non-monotonic (mean len 2.28–3.16), i.e. the positional defect is gone and
the remaining variance is content.

Fix

Rewriting the GGUF with two added keys fully restores long-context drafting:

dflash.attention.sliding_window          = 512        (u32)
dflash.attention.sliding_window_pattern  = [1,1,1,1,1,1]  (u32 array)

(The pattern array is required by the fork's loader once sliding_window is
set.) No tensor changes needed.

Suggestions

Re-publish the GGUF with the SWA metadata (this also affects CUDA users —
it is not a backend-specific issue).
Consider a loader-side warning or hard error in the fork when a dflash
model whose training config is all-SWA is loaded without window metadata —
the failure is silent and only visible as mysteriously bad long-context
acceptance.

Happy to share the patched-GGUF script and full
benchmark logs (2× AMD MI50, ROCm).

baranowskiadam

Poolside org about 7 hours ago

Hi @Dreadbyte , thanks for flagging it - we've discovered some issues with the uploaded GGUF configs yesterday and we'll be updating the checkpoints soon. I'll make sure to update this thread once they're fixed.

Dreadbyte

about 6 hours ago

Thank you!

baranowskiadam

Poolside org 3 minutes ago

Hi @Dreadbyte it's fixed! 🙏

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

DFlash GGUF missing sliding-window metadata -> draft acceptance collapses beyond short contexts

laguna-s-2.1-DFlash-BF16.gguf is missing sliding-window metadata → draft acceptance collapses to ~0% beyond short contexts

Summary

Observed impact

Fix

Suggestions

`laguna-s-2.1-DFlash-BF16.gguf` is missing sliding-window metadata → draft acceptance collapses to ~0% beyond short contexts