Image-Text-to-Text
GGUF
conversational

Model Details

This model is a gguf q4km format of google/gemma-4-31B-it generated by intel/auto-round algorithm using best recipe. Please refer to Section Generate the model for more details.

Please follow the license of the original model.

How To Use

Llama.cpp inference.

wget https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg
./llama-cli  -m gemma-4-31B-it-gguf-q4-k-m-AutoRound/gemma-4-31B-it-Q4_K_M.gguf -mm gemma-4-31B-it-gguf-q4-k-m-AutoRound/mmproj-model.gguf
> /image demo.jpeg

Loaded media from 'demo.jpeg'

> Describe this image.
[Start thinking]
Here's a thinking process that leads to the description:

1.  **Analyze the Request:** The user wants a description of the provided image.

2.  **Initial Scan/Overall Impression:**
    *   **Subject:** A woman and a dog.
    *   **Setting:** A sandy beach at sunset/sunrise.
    *   **Action:** The dog is giving the woman its paw.
    *   **Mood:** Peaceful, affectionate, warm, bonding.

3.  **Detailed Breakdown (Deconstructing the image):**

    *   **The People/Animals (Subjects):**
        *   *The Dog:* A light-brown/golden Labrador Retriever (or similar breed). It's sitting and lifting its right front paw. It's wearing a blue/patterned harness with a red leash attached, lying on the sand.
        *   *The Woman:* A young woman with long brown hair. She's wearing a blue and white plaid (flannel) shirt and dark pants. She's sitting cross-legged (or with one knee up) on the sand, facing the dog. She has a gentle smile on her face.

    *   **The Action/Interaction:**
        *   The dog has its paw extended, and the woman is gently holding it. This is a classic "shake" or "give paw" behavior, symbolizing a bond and training.

    *   **The Environment (Setting):**
        *   *Foreground:* Pale, textured sand with some footprints and shadows.
        *   *Midground:* The interaction between the woman and dog.
        *   *Background:* The ocean with a gentle wave breaking.
        *   *Sky:* Bright, pale, almost washed out due to the light source (likely the sun low on the horizon).

    *   **Lighting and Color Palette:**
        *   *Lighting:* Golden hour light. The sun is low, coming from the right, creating a warm glow and long shadows. There's a bit of lens flare/overexposure on the right side.
        *   *Colors:* Muted and warm. Tans/beiges of the sand and dog, soft blues of the harness and shirt, white/grey of the ocean foam, and a bright white/yellow sky.

4.  **Synthesizing into a Narrative Description (Drafting):**

    *   *Attempt 1 (Basic):* A woman is sitting on a beach with her golden retriever. The dog is giving her its paw. The woman is wearing a plaid shirt. It looks like sunset. The ocean is in the background.

    *   *Attempt 2 (More descriptive):* A heartwarming scene on a sandy beach during the golden hour. A young woman with long brown hair, wearing a blue and white plaid shirt, sits on the sand facing a light-brown Labrador. The dog, wearing a blue harness, is sitting up and offering its paw to the woman, who holds it
gently. In the background, soft waves are crashing on the shore under a bright, hazy sky.

    *   *Attempt 3 (Polishing for flow and detail - Final Version):* (This aligns with the final output provided to the user). I'll organize it logically: overall scene $\rightarrow$ subjects/action $\rightarrow$ background/lighting.

5.  **Final Review:** Does the description accurately reflect everything in the image? Yes. Is the tone appropriate? Yes. Is it detailed enough without being overly wordy? Yes.
[End thinking]

A heartwarming image captures a quiet moment of bonding between a woman and her dog on a sandy beach during the "golden hour" of sunrise or sunset.

In the center of the frame, a young woman with long brown hair is sitting on the pale sand. She is wearing a blue and white plaid flannel shirt and dark trousers. Facing her is a light-brown Labrador Retriever, who is sitting upright and lifting its right front paw into the woman's hand in a "shake" gesture. The dog i
s wearing a blue patterned harness with a red leash lying on the sand beside it.

The background features the ocean, with a gentle white wave breaking on the shore. The sky is bright and hazy, with the warm, glowing light of the low sun coming from the right, casting soft shadows across the textured sand and creating a peaceful, affectionate atmosphere.

Generate the model

Here is the sample command to reproduce the model

auto_round_best google/gemma-4-31B-it/  --format "gguf:q4_k_m" --output_dir tmp_autoround --enable_alg_ext

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

Downloads last month
-
GGUF
Model size
31B params
Architecture
gemma4
Hardware compatibility
Log In to add your hardware

4-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Intel/gemma-4-31B-it-gguf-q4-k-m-AutoRound

Quantized
(276)
this model

Dataset used to train Intel/gemma-4-31B-it-gguf-q4-k-m-AutoRound

Paper for Intel/gemma-4-31B-it-gguf-q4-k-m-AutoRound