Instructions to use tanlaan/obelisk-49l-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 tanlaan/obelisk-49l-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 tanlaan/obelisk-49l-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf tanlaan/obelisk-49l-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 tanlaan/obelisk-49l-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf tanlaan/obelisk-49l-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 tanlaan/obelisk-49l-gguf:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf tanlaan/obelisk-49l-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 tanlaan/obelisk-49l-gguf:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf tanlaan/obelisk-49l-gguf:Q4_K_M

Use Docker

docker model run hf.co/tanlaan/obelisk-49l-gguf:Q4_K_M

LM Studio
Jan

vLLM

How to use tanlaan/obelisk-49l-gguf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tanlaan/obelisk-49l-gguf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tanlaan/obelisk-49l-gguf",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/tanlaan/obelisk-49l-gguf:Q4_K_M

Ollama
How to use tanlaan/obelisk-49l-gguf with Ollama:
```
ollama run hf.co/tanlaan/obelisk-49l-gguf:Q4_K_M
```

Unsloth Studio

How to use tanlaan/obelisk-49l-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 tanlaan/obelisk-49l-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 tanlaan/obelisk-49l-gguf to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for tanlaan/obelisk-49l-gguf to start chatting

Docker Model Runner
How to use tanlaan/obelisk-49l-gguf with Docker Model Runner:
```
docker model run hf.co/tanlaan/obelisk-49l-gguf:Q4_K_M
```

Lemonade

How to use tanlaan/obelisk-49l-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull tanlaan/obelisk-49l-gguf:Q4_K_M

Run and chat with the model

lemonade run user.obelisk-49l-gguf-Q4_K_M

List all available models

lemonade list

Atomic Chat

obelisk-49l-gguf

GGUF exports for the tanlaan/obelisk-49l CPT repair checkpoint.

Source model

Dense HF model: tanlaan/obelisk-49l
Source checkpoint line: repaired Granite 49L branch with staged Dolmino ingredient1 continual pretraining
Effective CPT depth for this export: ~64M tokens
Sequence length during CPT: 8192

Files

obelisk-49l-f16.gguf
- unquantized reference GGUF
- size: about 1.1 GB
obelisk-49l-Q4_K_M.gguf
- intended for mobile / PocketPal / lighter local runtimes
- size: about 332 MB
obelisk-49l-Q8_0.gguf
- higher-quality reference quantization
- size: about 568 MB

Evaluation snapshot

Checkpoint	ARC	HellaSwag	MMLU Mini	TruthfulQA MC2	Winogrande	GSM8K Flex	GSM8K Strict	5-task Mean
Base original	0.36	0.515	0.31	0.4200	0.60	0.18	n/a	0.3975
Recovered base	0.35	0.545	0.315	0.3955	0.55	0.20	0.20	0.4311
`~64M effective @ 8192` continuation	0.38	0.52	0.255	0.4422	0.60	0.27	0.26	0.4394

Notes

This repo contains quantized GGUF exports derived from the dense Hugging Face checkpoint.
The HF source repo is the better choice if you want the original transformers model artifacts.
The Q4_K_M file is the first file to try for PocketPal and similar phone runtimes.

Downloads last month: 15

GGUF

Model size

0.6B params

Architecture

granitehybrid

Hardware compatibility

4-bit

8-bit

16-bit

Model tree for tanlaan/obelisk-49l-gguf

Base model

tanlaan/obelisk-49l

Quantized

(1)

this model