OPEA/gemma-3-12b-it-AutoRound-gguf-q4-0

Model Details

This model is an int4 model with group_size 32 and symmetric quantization of google/gemma-3-12b-it generated by intel/auto-round algorithm.

Please follow the license of the original model.

How To Use

Requirements

Please follow the Build llama.cpp locally to install the necessary dependencies.

INT4 Inference

This model has vision capabilities, more details here: https://github.com/ggml-org/llama.cpp/pull/12344

After building with Gemma 3 clip support, run the following command:

>>> wget https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg
>>> ./build/bin/llama-gemma3-cli -m gemma-3-12b-it-12B-Q4_0.gguf --mmproj mmproj-gemma-3-12b-it-12B-Q4_0.gguf

Output:

 Running in chat mode, available commands:
   /image <path>    load an image
   /clear           clear the chat history
   /quit or /exit   exit the program
> /image bee.jpg                
Encoding image bee.jpg

> Describe this image in detail.
Here's a detailed description of the image:

**Overall Impression:**

The image is a close-up shot of a cluster of pink cosmos flowers with a bumblebee actively foraging on one of them. The scene has a natural, slightly wild feel.

**Flowers:**

*   **Type:** Cosmos flowers, identifiable by their characteristic daisy-like shape.
*   **Color:** Primarily pink, with varying shades from a lighter, almost pastel pink to a slightly deeper, richer pink.
*   **Condition:** Some flowers are in full bloom, while others are past their prime, showing signs of wilting and drying. Some have brown, dried seed heads.
*   **Arrangement:** The flowers are clustered together, creating a dense, somewhat chaotic arrangement.

**Bumblebee:**

*   **Position:** A bumblebee is positioned on the central, most prominent pink cosmos flower. It appears to be actively feeding, likely collecting nectar or pollen.
*   **Appearance:** The bee has the classic fuzzy, black and yellow striped pattern of a bumblebee. Its legs are visible, and it seems to be deeply embedded within the flower.

**Background:**

*   **Foliage:** The background is filled with green foliage, including large, broad leaves and smaller, more delicate stems.
*   **Other Flowers:** There are hints of other flowers in the background, including a single red flower.
*   **Depth of Field:** The depth of field is shallow, meaning the foreground flowers and the bee are in sharp focus, while the background is blurred, drawing attention to the main subject.

**Lighting and Composition:**

*   **Lighting:** The lighting appears to be natural, likely daylight.
*   **Composition:** The composition is well-balanced, with the central flower and bee serving as the focal point. The surrounding flowers and foliage create a visually interesting and natural frame.

**Overall Tone:**

The image evokes a sense of natural beauty, vibrancy, and the busy activity of pollinators in a garden setting.

Generate the model

Here is the sample command to reproduce the model.

pip install git+https://github.com/intel/auto-round.git@main
auto-round-mllm \
--model google/gemma-3-12b-it \
--device 0 \
--bits 4 \
--group_size 32 \
--batch_size 1 \
--gradient_accumulate_steps 8 \
--format 'gguf:q4_0' \
--output_dir "./tmp_autoround"

Ethical Considerations and Limitations

The model can produce factually incorrect output, and should not be relied on to produce factually accurate information. Because of the limitations of the pretrained model and the finetuning datasets, it is possible that this model could generate lewd, biased or otherwise offensive outputs.

Therefore, before deploying any applications of the model, developers should perform safety testing.

Caveats and Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model.

Here are a couple of useful links to learn more about Intel's AI software:

• Intel Neural Compressor link

Disclaimer

The license on this model does not constitute legal advice. We are not responsible for the actions of third parties who use this model. Please consult an attorney before using this model for commercial purposes.

Cite

@article{cheng2023optimize, title={Optimize weight rounding via signed gradient descent for the quantization of llms}, author={Cheng, Wenhua and Zhang, Weiwei and Shen, Haihao and Cai, Yiyang and He, Xin and Lv, Kaokao and Liu, Yi}, journal={arXiv preprint arXiv:2309.05516}, year={2023} }

arxiv github