Qualcomm AI Hub Community

Contribution Guidelines

Welcome to the community model repository! This Space is designed to host models contributed by users in a structured and transparent way. Below are the instructions and standards for contributing your model.

📁 Model Repository

Each model must be placed in its own repository. The folder should follow this structure:

<repo-name>/
    ├── LICENSE (optional, strongly recommended)
    ├── README.md (optional)
    ├── <model_name>_dlc_<precision>.zip
    ├── <model_name>_onnx_<precision>.zip
    ├── <model_name>_tflite_<precision>.zip
    ├── <model_name>_qnn_context_binary_<precision>_<chipset_name>.zip
    ├── <model_name>_genie_<precision>_<chipset_name>.zip
    └── <model_name>_onnxruntime_genai_<precision>_<chipset_name>.zip

• LICENSE: The license under which your model is shared.
• README.md: A markdown file describing your model, its purpose, usage, and any relevant metadata.
• Model files: Contains zip files for each runtime, precision and device (if the asset is device dependent). Mandatory to atleast contribute one zip file for a model.

Each zip file must include:

• The model file(s)
• tool-versions.yaml
• Other essential files required for on-device deployment

Example: ONNX and DLC Export When unzipped, an ONNX and DLC model folder might look like this:

<model_name>_dlc_<precision>/ 
├── tool_versions.yaml 
├── model.dlc
└── labels.json (optional)

<model_name>_onnx_<precision>/ 
├── tool_versions.yaml 
├── model.onnx
└── labels.json (optional)

Example: LLM Genie Export When unzipped, a Genie model folder might look like this:

<model_name>_genie_<precision>_<chipset-name>/ 
├── tool_versions.yaml 
├── model_part1.bin
├── model_part2.bin
├── model_part3.bin
├── tokenizer.yaml
├── htp_config.json
└── genie_config.json

🧰 tool_versions.yaml

To ensure reproducibility and correct deployment of model assets, a tool-versions.yaml file should be there in each zip. This file should specify the versions of tools used during compilation.

tool_versions:
  aihm_version: 0.43.0
  qairt: 2.34.0
  onnx_runtime: 1.20
  qairt: 2.38.0
  ...

✅ How to Contribute

• Export any model from AI Hub Models using the --zip-assets argument (minimum AI Hub Models package: 0.43.0).
• This ensures the model is saved in the required format in your specified output directory.
• If contributing multiple precisions or runtimes, export multiple times with the same output directory and --zip-assets argument.
• Upload the contents of the output directory to your new repository in this space.
• To contribute a new model, review the export.py script of any model in AI Hub Models.

⚠️ Disclaimer

This Space is a community-driven repository. The models hosted here are user contributions and:

• Are not verified by the organization or maintainers for correctness, safety, or performance.
• May contain errors, bugs, or limitations.
• Are moderated only for structural compliance, not for content quality.

The organization and maintainers do not take responsibility for the models or assets contributed here. Use them at your own discretion.

Thank you for contributing and helping build a rich ecosystem of models!