Instructions to use vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16 with MLX:
# Make sure mlx-vlm is installed # pip install --upgrade mlx-vlm from mlx_vlm import load, generate from mlx_vlm.prompt_utils import apply_chat_template from mlx_vlm.utils import load_config # Load the model model, processor = load("vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16") config = load_config("vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16") # Prepare input image = ["http://images.cocodataset.org/val2017/000000039769.jpg"] prompt = "Describe this image." # Apply chat template formatted_prompt = apply_chat_template( processor, config, prompt, num_images=1 ) # Generate output output = generate(model, processor, formatted_prompt, image) print(output) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Pi
How to use vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16 with Pi:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16"
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": "vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Hermes Agent new
How to use vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16 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 "vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16"
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 vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16
Run Hermes
hermes
Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16
MLX-VLM conversion of huihui-ai/Huihui-gemma-4-26B-A4B-it-abliterated.
Overview
- Format:
MLX-VLM - Precision:
bf16 - Size:
48G - Source model type:
Gemma4ForConditionalGeneration - Source pipeline:
any-to-any - Intended runtime:
mlx-vlmandLM Studio
Conversion Notes
- Converted for local Apple Silicon inference with
mlx-vlm - Kept in MLX-VLM multimodal layout for image-text generation
- Includes local config compatibility fixes required during conversion/debugging
Validation
Local checks on Apple Silicon:
- model loading in
mlx-vlm:passed - local conversion / quantization:
passed - text generation smoke test:
mixed - image generation path smoke test:
mixed - lm studio 0.4.11+1 loading:
not supported yet - notes:
Conversion finished successfully, but Gemma 4 runtime behavior still depends on current mlx-vlm and LM Studio support. In local checks, the models load in mlx-vlm and can emit tokens, but output quality still needs tuning, and LM Studio 0.4.11+1 currently refuses to load Gemma 4.
Files
Important files in this repo:
config.jsongeneration_config.jsonchat_template.jinjaprocessor_config.jsontokenizer.jsontokenizer_config.jsonmodel.safetensors.index.jsonmodel-*.safetensors
Usage
Text generation
mlx_vlm.generate \
--model /path/to/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16 \
--prompt "Describe local inference in one short sentence." \
--max-tokens 128 \
--temperature 1.0 \
--trust-remote-code
Image prompt
mlx_vlm.generate \
--model /path/to/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16 \
--image /path/to/example.png \
--prompt "Describe this image." \
--max-tokens 128 \
--temperature 1.0 \
--trust-remote-code
LM Studio
LM Studio 0.4.11+1 detects these repos but currently refuses to load Gemma 4 with ValueError: Gemma 4 support is not ready yet, stay tuned!. Use mlx-vlm for now until LM Studio adds native Gemma 4 support.
Notes
This repo reflects a local conversion workflow and validation pass on Apple Silicon. Behavior can vary with mlx-vlm version, sampling parameters, and prompt style.
- Downloads last month
- 162
Quantized
Model tree for vanch007/Huihui-gemma-4-26B-A4B-it-abliterated-mlx-bf16
Base model
google/gemma-4-26B-A4B