kangjirm/SmolVLM2-500M-Video-Instruct-Chakra-Q8_0.gguf

SmolVLM2-500M-Video-Instruct-Chakra-Q8_0

Fine-tuned SmolVLM2 model for fortune-telling and UI element description.

Model Description

This is a quantized (Q8_0) version of SmolVLM2-500M-Video-Instruct, fine-tuned for:

• Fortune-telling descriptions: Providing contextual fortune interpretations based on visual elements
• UI element classification: Describing UI components and their purposes in AR applications

Files

• smolvlm_element_descriptive_q8.gguf (437MB): Main vision-language model
• mmproj-SmolVLM2-500M-Video-Instruct-Q8_0.gguf (109MB): Multi-modal projection layer

Use Case

This model is specifically designed for the Fortuna AR application, which:

• Detects objects in AR environment
• Provides fortune-telling interpretations based on detected elements
• Describes UI components for enhanced user experience

Quantization

• Format: GGUF (Q8_0 quantization)
• Compatibility: llama.cpp
• Platform: Optimized for mobile deployment (Android)

Usage

val modelPath = "path/to/smolvlm_element_descriptive_q8.gguf"
val mmprojPath = "path/to/mmproj-SmolVLM2-500M-Video-Instruct-Q8_0.gguf"

Base Model

Based on SmolVLM2-500M-Video-Instruct

License

This model is licensed under Apache 2.0. Please refer to the base model's license terms for more details.

Citation

@misc{smolvlm2-fortuna-finetune,
  title={SmolVLM2 Fine-tuned for Fortune-telling and UI Description},
  author={Fortuna Team},
  year={2025}
}