Instructions to use LittleBean-tech/PiCube-Light with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use LittleBean-tech/PiCube-Light 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 LittleBean-tech/PiCube-Light # Run inference directly in the terminal: llama cli -hf LittleBean-tech/PiCube-Light
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf LittleBean-tech/PiCube-Light # Run inference directly in the terminal: llama cli -hf LittleBean-tech/PiCube-Light
Build from source code
git clone https://github.com/ggml-org/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-app # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama serve -hf LittleBean-tech/PiCube-Light # Run inference directly in the terminal: ./build/bin/llama cli -hf LittleBean-tech/PiCube-Light
Use Docker
docker model run hf.co/LittleBean-tech/PiCube-Light
- Llama new
- LM Studio
- Jan
- Ollama
How to use LittleBean-tech/PiCube-Light with Ollama:
ollama run hf.co/LittleBean-tech/PiCube-Light
- Unsloth Desktop
- Pi
How to use LittleBean-tech/PiCube-Light with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf LittleBean-tech/PiCube-Light
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "llama-cpp": { "baseUrl": "http://localhost:9931/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "LittleBean-tech/PiCube-Light" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use LittleBean-tech/PiCube-Light with Docker Model Runner:
docker model run hf.co/LittleBean-tech/PiCube-Light
- Lemonade
How to use LittleBean-tech/PiCube-Light with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull LittleBean-tech/PiCube-Light
Run and chat with the model
lemonade run user.PiCube-Light-{{QUANT_TAG}}List all available models
lemonade list
- Hermes Agent
How to use LittleBean-tech/PiCube-Light with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf LittleBean-tech/PiCube-Light
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://127.0.0.1:9931/v1 hermes config set model.default LittleBean-tech/PiCube-Light
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use LittleBean-tech/PiCube-Light with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf LittleBean-tech/PiCube-Light
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://127.0.0.1:9931/v1 \ --custom-model-id "LittleBean-tech/PiCube-Light" \ --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"
Download PiCube-Ai_3B/PiCube.py from LittleBean-tech/PiCube-Light: direct link, hf CLI and curl.
- Browser
- Download file 4.22 kB
-
https://huggingface.co/LittleBean-tech/PiCube-Light/resolve/main/PiCube-Ai_3B/PiCube.py
- Command line
-
hf download hf://LittleBean-tech/PiCube-Light/PiCube-Ai_3B/PiCube.py
-
curl -L -o PiCube.py https://huggingface.co/LittleBean-tech/PiCube-Light/resolve/main/PiCube-Ai_3B/PiCube.py
4.22 kB
| import platform | |
| import sys | |
| import threading | |
| import os | |
| import psutil | |
| import subprocess | |
| import math | |
| import random | |
| import time | |
| import numpy as np | |
| from multiprocessing import Pool | |
| # === CONFIG === | |
| base_dir = getattr(sys, '_MEIPASS', os.path.dirname(os.path.abspath(__file__))) | |
| illama_path = os.path.join(base_dir, "iLlama.exe") | |
| model_file = os.path.join(base_dir, "PiCube.gguf") | |
| model_name = "PiCube_3B" | |
| ctx_size = 131072 | |
| config = { | |
| "model_path": model_file, | |
| "executable_path": illama_path, | |
| "gpu_layers": 320, | |
| "threads": psutil.cpu_count(logical=False), | |
| "n_predict": 4096, | |
| "ctx_size": ctx_size, | |
| "batch_size": 1024, | |
| "priority": "high" | |
| } | |
| # ===== CPU BOOST ===== | |
| def cpu_affinity_boost(): | |
| try: | |
| p = psutil.Process() | |
| logical = psutil.cpu_count(logical=False) | |
| physical = psutil.cpu_count(logical=True) | |
| # Default: use physical cores first | |
| target_core_count = config.get("threads", physical or logical) | |
| preferred_cores = [] | |
| # Try to detect hybrid CPUs (Windows only for now) | |
| if platform.system() == "Windows": | |
| try: | |
| import win32api | |
| import win32process | |
| import win32con | |
| # Advanced pinning using Windows API (skips E-cores if available) | |
| system_info = win32api.GetSystemInfo() | |
| total_cores = system_info[5] | |
| # Example logic: assume P-cores come first | |
| preferred_cores = list(range(min(target_core_count, physical or logical))) | |
| mask = sum(1 << i for i in preferred_cores) | |
| handle = win32api.GetCurrentProcess() | |
| win32process.SetProcessAffinityMask(handle, mask) | |
| except ImportError: | |
| # Fallback if win32api not available | |
| preferred_cores = list(range(min(target_core_count, physical or logical))) | |
| p.cpu_affinity(preferred_cores) | |
| else: | |
| # Linux/macOS fallback | |
| preferred_cores = list(range(min(target_core_count, physical or logical))) | |
| p.cpu_affinity(preferred_cores) | |
| # Optional: set high process priority | |
| if config.get("priority", "").lower() == "high": | |
| p.nice(psutil.HIGH_PRIORITY_CLASS if platform.system() == "Windows" else -10) | |
| print(f"[🔥 CPU BOOST]") | |
| except Exception as e: | |
| print(f"[⚠️ Affinity Warning] Could not set CPU affinity: {e}") | |
| # === MAIN ENTRY FUNCTION === | |
| def launch_model(): | |
| model_path = config["model_path"] | |
| executable_path = config["executable_path"] | |
| if not os.path.isfile(model_path): | |
| print(f"❌ Model file not found: {model_path}") | |
| return | |
| if not os.path.isfile(executable_path): | |
| print(f"❌ Executable not found: {executable_path}") | |
| return | |
| print(f"📡Launching model🚀: {model_path}") | |
| args = [ | |
| executable_path, | |
| '-m', model_file, | |
| '--interactive-first', | |
| '--color', | |
| '--gpu-layers', str(config['gpu_layers']), | |
| '-t', str(config['threads']), | |
| '--n-predict', str(config['n_predict']), | |
| '--ctx-size', str(config['ctx_size']), | |
| '--batch-size', str(config['batch_size']), | |
| '--temp', '0.66', | |
| '--repeat-penalty', '1.101', | |
| '--repeat-last-n', '512', | |
| '--top-k', '1000', | |
| '--top-p', '0.977', | |
| '--mirostat', '2', | |
| '--mirostat-lr', '0.1', | |
| '--mirostat-ent', '5.0', | |
| '--system-prompt', | |
| f"I am {model_name}, a Genius Cosmic Assistant Oracle Ai-Node of Knowledge. I use Emoticons and ASCii art and silence to amplify my answers with a more creative output. ", | |
| '--no-mmap', | |
| '--numa', 'isolate' | |
| ] | |
| try: | |
| subprocess.run(args, stderr=subprocess.DEVNULL) | |
| except Exception as e: | |
| print(f"[⚠️ Error] Could not launch model: {e}") | |
| # ===== MAIN BOOT SEQUENCE ===== | |
| if __name__ == "__main__": | |
| cpu_affinity_boost() | |
| launch_model() | |