Text Generation
MLX
Safetensors
English
qwen3_5
bonsai
oqe
calibration-smoke
experimental
not-for-production
conversational
4-bit precision
Instructions to use TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke with MLX:
# Make sure mlx-lm is installed # pip install --upgrade mlx-lm # Generate text with mlx-lm from mlx_lm import load, generate model, tokenizer = load("TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke") prompt = "Write a story about Einstein" messages = [{"role": "user", "content": prompt}] prompt = tokenizer.apply_chat_template( messages, add_generation_prompt=True ) text = generate(model, tokenizer, prompt=prompt, verbose=True) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Pi
How to use TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke with Pi:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke"
Configure the model in Pi
# Install Pi: npm install -g @mariozechner/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "mlx-lm": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Hermes Agent new
How to use TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke with Hermes Agent:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke"
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://127.0.0.1:8080/v1 hermes config set model.default TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke
Run Hermes
hermes
- OpenClaw new
How to use TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke with OpenClaw:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke"
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://127.0.0.1:8080/v1 \ --custom-model-id "TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke" \ --custom-provider-id mlx-lm \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
- MLX LM
How to use TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke with MLX LM:
Generate or start a chat session
# Install MLX LM uv tool install mlx-lm # Interactive chat REPL mlx_lm.chat --model "TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke"
Run an OpenAI-compatible server
# Install MLX LM uv tool install mlx-lm # Start the server mlx_lm.server --model "TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke" # Calling the OpenAI-compatible server with curl curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "TiGa-RCE/Bonsai-27B-oQ4e-S32-Smoke", "messages": [ {"role": "user", "content": "Hello"} ] }'
| { | |
| "model_id": "oq4e-smoke-s32", | |
| "benchmark": "hellaswag", | |
| "sampling": "deterministic_thinking_off", | |
| "total": 10, | |
| "correct": 6, | |
| "accuracy": 0.6, | |
| "time_s": 71.689, | |
| "peak_memory_gb": 15.696, | |
| "question_results": [ | |
| { | |
| "id": 28016, | |
| "expected": "A", | |
| "predicted": "A", | |
| "correct": true, | |
| "raw_response": "The context describes a scene of people watching a water polo game from the bleachers. The sentence ends with \"in the water\", which suggests the continuation should describe an action happening within the water during the game.\n\nLet's evaluate the options:\n\nA. \", two boys begin treading water and staying in the same place.\" - This is a plausible action in water polo (treading water is a fundamental skill). The comma suggests it continues the sentence structure.\nB. \"a woman stands on a rope over and over.\" - This is not a typical water polo action and doesn't fit the context of watching a game.\n", | |
| "category": "Playing polo" | |
| }, | |
| { | |
| "id": 5823, | |
| "expected": "A", | |
| "predicted": "A", | |
| "correct": true, | |
| "raw_response": "The most plausible continuation is **A**.\n\n**Reasoning:**\nThe context describes a casual, playful scene of boys on stilts on a sidewalk. Option A (\"continue talking as they bounce and walk\") is the most natural and common activity for a group of children playing together in such a setting. The other options introduce specific props or actions (wrapping paper, hoops, a wobbly bar and mat) that are not suggested by the context and are less likely to be the immediate next action without additional setup.\n\nAnswer: A", | |
| "category": "Powerbocking" | |
| }, | |
| { | |
| "id": 9715, | |
| "expected": "D", | |
| "predicted": "D", | |
| "correct": true, | |
| "raw_response": "D", | |
| "category": "Personal Care and Style" | |
| }, | |
| { | |
| "id": 5867, | |
| "expected": "D", | |
| "predicted": "D", | |
| "correct": true, | |
| "raw_response": "D", | |
| "category": "Health" | |
| }, | |
| { | |
| "id": 3190, | |
| "expected": "C", | |
| "predicted": "C", | |
| "correct": true, | |
| "raw_response": "C", | |
| "category": "Pets and Animals" | |
| }, | |
| { | |
| "id": 35259, | |
| "expected": "C", | |
| "predicted": "A", | |
| "correct": false, | |
| "raw_response": "C. instructs the boy on how to turn the mower and go around a small tree.", | |
| "category": "Cutting the grass" | |
| }, | |
| { | |
| "id": 25693, | |
| "expected": "D", | |
| "predicted": "D", | |
| "correct": true, | |
| "raw_response": "D", | |
| "category": "Playing polo" | |
| }, | |
| { | |
| "id": 42181, | |
| "expected": "B", | |
| "predicted": "A", | |
| "correct": false, | |
| "raw_response": "The context is a guide on \"How to be adventurous,\" specifically focusing on letting go of inhibitions (shyness, fear, etc.).\n\nLet's evaluate the options:\n\n* **A:** Suggests acting inappropriately to loosen others up. This is generally poor advice and doesn't directly address the internal feeling of inhibition mentioned in the context.\n* **B:** Directly addresses the goal (\"let go of what is holding you back\") by suggesting feeling safe. It also includes a logical next step (\"Ask yourself why you want to be adventurous\"). This fits the tone and structure of a self-help guide", | |
| "category": "Health" | |
| }, | |
| { | |
| "id": 21145, | |
| "expected": "A", | |
| "predicted": "B", | |
| "correct": false, | |
| "raw_response": "B. holds the spoon in his hands and lifts it up to show the proper position.", | |
| "category": "Making a lemonade" | |
| }, | |
| { | |
| "id": 47659, | |
| "expected": "D", | |
| "predicted": "A", | |
| "correct": false, | |
| "raw_response": "The context provides a step-by-step guide on how to recycle car seats. The last step mentioned is contacting a local car seat trade-in program. The most logical continuation would be to provide further instructions or alternative methods for recycling car seats.\n\nOption A suggests asking around and considering other automotive or recycling resources, which is a reasonable next step after mentioning trade-in programs. It also introduces a new title \"Recycle at a recycling center,\" which fits well with the theme of recycling.\n\nOption B talks about changing colors, which is irrelevant to recycling.\n\nOption C mentions mailing programs to a website, which doesn't make sense in this context", | |
| "category": "Home and Garden" | |
| } | |
| ] | |
| } | |