Instructions to use YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco")
model = AutoModelForCausalLM.from_pretrained("YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco

SGLang

How to use YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco with Docker Model Runner:
```
docker model run hf.co/YeonwooSung/llama-3.1-8B-Galore-openassistant-guanaco
```

Model Card for Model ID

Model Details

Training Details

Training Data

timdettmers/openassistant-guanaco

Training Procedure

Trained with SFTTrainer with Galore quantization.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	18.07
IFEval (0-Shot)	26.35
BBH (3-Shot)	31.44
MATH Lvl 5 (4-Shot)	4.83
GPQA (0-shot)	6.71
MuSR (0-shot)	14.58
MMLU-PRO (5-shot)	24.52

Downloads last month: 29

Safetensors

Model size

8B params

Tensor type

BF16

Evaluation results

strict accuracy on IFEval (0-Shot)
Open LLM Leaderboard

26.350
normalized accuracy on BBH (3-Shot)
Open LLM Leaderboard

31.440
exact match on MATH Lvl 5 (4-Shot)
Open LLM Leaderboard

4.830
acc_norm on GPQA (0-shot)
Open LLM Leaderboard

6.710
acc_norm on MuSR (0-shot)
Open LLM Leaderboard

14.580
accuracy on MMLU-PRO (5-shot)
test set Open LLM Leaderboard

24.520