Instructions to use Lucebox/Laguna-XS.2-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Lucebox/Laguna-XS.2-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Lucebox/Laguna-XS.2-GGUF",
	filename="laguna-xs2-Q4_K_M.gguf",
)

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

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use Lucebox/Laguna-XS.2-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Lucebox/Laguna-XS.2-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Lucebox/Laguna-XS.2-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Lucebox/Laguna-XS.2-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Lucebox/Laguna-XS.2-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 Lucebox/Laguna-XS.2-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf Lucebox/Laguna-XS.2-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 Lucebox/Laguna-XS.2-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Lucebox/Laguna-XS.2-GGUF:Q4_K_M

Use Docker

docker model run hf.co/Lucebox/Laguna-XS.2-GGUF:Q4_K_M

LM Studio
Jan
Ollama
How to use Lucebox/Laguna-XS.2-GGUF with Ollama:
```
ollama run hf.co/Lucebox/Laguna-XS.2-GGUF:Q4_K_M
```

Unsloth Studio new

How to use Lucebox/Laguna-XS.2-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 Lucebox/Laguna-XS.2-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 Lucebox/Laguna-XS.2-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Lucebox/Laguna-XS.2-GGUF to start chatting

Docker Model Runner
How to use Lucebox/Laguna-XS.2-GGUF with Docker Model Runner:
```
docker model run hf.co/Lucebox/Laguna-XS.2-GGUF:Q4_K_M
```

Lemonade

How to use Lucebox/Laguna-XS.2-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Lucebox/Laguna-XS.2-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Laguna-XS.2-GGUF-Q4_K_M

List all available models

lemonade list

davide221 commited on 2 days ago

Commit

b42efde

verified ·

1 Parent(s): 7ed26b3

Upload README.md with huggingface_hub

Browse files

Files changed (1) hide show

README.md +52 -0

README.md ADDED Viewed

	@@ -0,0 +1,52 @@

+---
+license: apache-2.0
+base_model: poolside/Laguna-XS.2
+tags:
+- gguf
+- llama.cpp
+- moe
+- code
+- quantized
+---
+# Laguna-XS.2 GGUF (BF16 + Q4_K_M)
+GGUF conversions of [poolside/Laguna-XS.2](https://huggingface.co/poolside/Laguna-XS.2) for use with [llama.cpp](https://github.com/ggerganov/llama.cpp).
+## Files
+| File | Quant | Size | BPW | Notes |
+|------|-------|------|-----|-------|
+| `laguna-xs2-bf16.gguf` | BF16 | 62.3 GiB | 16.01 | reference, identical math to HF transformers fp/bf16 |
+| `laguna-xs2-Q4_K_M.gguf` | Q4_K_M | 18.88 GiB | 4.85 | imatrix-calibrated, fits a single 24 GB GPU |
+| `laguna-xs2.imatrix` | imatrix | 180 MB | — | Bartowski calibration_datav3 (134 chunks, 68608 tokens) |
+## Quality
+| Metric | BF16 | Q4_K_M | Δ |
+|--------|------|--------|---|
+| Perplexity (Bartowski v3, 20×512) | 10.7594 ± 0.522 | 11.2854 ± 0.553 | +4.9% |
+Imatrix calibration uses Bartowski `calibration_datav3.txt` (multilingual + code mix), the same corpus Unsloth-distributed quants use.
+## Required llama.cpp patch
+Laguna-XS.2 is a NEW architecture (`LLM_ARCH_LAGUNA`) not present in upstream llama.cpp. Loading these GGUFs requires a llama.cpp build with the LAGUNA arch added. Patches available at: https://github.com/your-org/lucebox-hub (see `dflash/deps/llama.cpp/src/models/laguna.cpp` and related changes).
+Architecture summary:
+- 40 layers hybrid iSWA: pattern (full, sw, sw, sw) × 10, sliding window 512
+- Per-layer head count: 48 (full) / 64 (sliding); 8 KV heads always
+- Q-norm + K-norm RMSNorm at head_dim level
+- Per-head softplus attention gate
+- RoPE per layer-type: full layers YaRN (theta=500K, factor=32, partial_rotary=0.5); sliding default (theta=10K, partial_rotary=1.0)
+- 256 experts, top-8, sigmoid router with score-correction bias, sum-normalize, scale=2.5
+- Always-on shared expert (intermediate=512)
+- Dense layer 0, sparse MoE layers 1–39
+## Quick test
+```bash
+./llama-simple -m laguna-xs2-Q4_K_M.gguf -ngl 99 -n 128 "def fibonacci(n):"
+```
+Tested on RTX 3090 24GB and A100 80GB. Inference @ 154 tok/s on A100 SXM Q4_K_M.