|
Download README.md from ayanoby/hexanima-h3: direct link, hf CLI and curl.
- Browser
- Download file 2.15 kB
-
https://huggingface.co/ayanoby/hexanima-h3/resolve/main/README.md
- Command line
-
hf download hf://ayanoby/hexanima-h3/README.md
-
curl -L -o README.md https://huggingface.co/ayanoby/hexanima-h3/resolve/main/README.md
2.15 kB
| license: other | |
| license_name: minimax-h3-image-community-license-agreement | |
| license_link: LICENSE | |
| base_model: MiniMaxAI/MiniMax-H3-Image | |
| base_model_relation: quantized | |
| # MiniMax-H3-Image for HexAnima (Snapdragon NPU pack) | |
| This is a **modified** (repacked) version of | |
| [MiniMaxAI/MiniMax-H3-Image](https://huggingface.co/MiniMaxAI/MiniMax-H3-Image) | |
| for the HexAnima Android app (Hexagon V69 NPU). It is a Model Derivative in the sense of the license. | |
| ## Modifications (Section III.2 of the license) | |
| All weight files in this repository were modified from the originals: | |
| - `model/`: the INT8-ConvRot transformer from | |
| [unsloth/MiniMax-H3-Image-FP8](https://huggingface.co/unsloth/MiniMax-H3-Image-FP8) repacked into HMX V69 | |
| int8 tiles ("compact q8": one fp16 scale per tile column, rotation matrix `H256`), one file per block. | |
| - `te/`: the INT8-ConvRot Qwen3-VL text encoder (layers 0..49 + embedding) from the same source, repacked the same way. | |
| - `vae/vae_h3.bin`: the VAE decoder converted to a Qualcomm QNN HTP context binary (fp16). | |
| - `lora/`: the third-party pDMD 2-NFE LoRA | |
| ([akhaliq/minimax-h3-image-pdmd-2step](https://huggingface.co/akhaliq/minimax-h3-image-pdmd-2step), rank 128) | |
| repacked into tiles; it is applied on top of the transformer and is subject to the same license. | |
| - `tok_*.txt`: the Qwen3-VL tokenizer files in a plain text form. | |
| Nothing here is meant to be used outside the Applicable Territory. | |
| Excluded territories: the United States, the European Union, the United Kingdom, South Korea | |
| (see `LICENSE`, Section I). Outputs may not be used to improve any other AI model (Section V.3). | |
| Acceptable Use Policy: Exhibit A of `LICENSE`. | |
| ## Layout | |
| The HexAnima app (1.6.0 or newer) downloads this pack by itself from its model catalog into `files/h3/` (`model/`, `te/`, `lora/`, `vae/`, `tok_*.txt`, the licence texts) and checks every file against `manifest.json` (size and sha256). It needs about 48 GB of free storage and a Snapdragon with Hexagon V69 (8 Gen 1 / 8+ Gen 1). To install by hand, copy the folders as they are to `files/h3/` and create the empty marker `.installed`. | |
| Powered by MiniMax-H3-Image. | |