Instructions to use macadeliccc/piccolo-8x7b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use macadeliccc/piccolo-8x7b with Transformers:

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

pipe = pipeline("text-generation", model="macadeliccc/piccolo-8x7b")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("macadeliccc/piccolo-8x7b")
model = AutoModelForCausalLM.from_pretrained("macadeliccc/piccolo-8x7b")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use macadeliccc/piccolo-8x7b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "macadeliccc/piccolo-8x7b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "macadeliccc/piccolo-8x7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/macadeliccc/piccolo-8x7b

SGLang

How to use macadeliccc/piccolo-8x7b 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 "macadeliccc/piccolo-8x7b" \
    --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": "macadeliccc/piccolo-8x7b",
		"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 "macadeliccc/piccolo-8x7b" \
        --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": "macadeliccc/piccolo-8x7b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use macadeliccc/piccolo-8x7b with Docker Model Runner:
```
docker model run hf.co/macadeliccc/piccolo-8x7b
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Piccolo-8x7b

In loving memory of my dog Klaus (Piccolo)

~ Piccolo (Italian): the little one ~

Based on mlabonne/NeuralBeagle-7b Quants are available here

Code Example

Inference and Evaluation colab available here

from transformers import AutoModelForCausalLM, AutoTokenizer

def generate_response(prompt):
    """
    Generate a response from the model based on the input prompt.
    Args:
    prompt (str): Prompt for the model.

    Returns:
    str: The generated response from the model.
    """
    inputs = tokenizer(prompt, return_tensors="pt")
    outputs = model.generate(**inputs, max_new_tokens=256, eos_token_id=tokenizer.eos_token_id, pad_token_id=tokenizer.pad_token_id)

    response = tokenizer.decode(outputs[0], skip_special_tokens=True)

    return response

model_id = "macadeliccc/piccolo-8x7b"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id,load_in_4bit=True)

prompt = "What is the best way to train Cane Corsos?"

print("Response:")
print(generate_response(prompt), "\n")

The model is capable of quality code, math, and logical reasoning. Try whatever questions you think of.

Example output

Evaluations

https://huggingface.co/datasets/open-llm-leaderboard/details_macadeliccc__piccolo-8x7b

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	72.80
AI2 Reasoning Challenge (25-Shot)	69.62
HellaSwag (10-Shot)	86.98
MMLU (5-Shot)	64.13
TruthfulQA (0-shot)	64.17
Winogrande (5-shot)	79.87
GSM8k (5-shot)	72.02