| # Component licenses |
|
|
| This repository contains derivative weights with separate component licenses. |
| The complete model bundle is not licensed solely under Apache-2.0 or MIT. |
|
|
| | Component | Upstream revision | Terms | |
| | --- | --- | --- | |
| | `denoiser/`, `octreeDecoder/`, `gaussianDecoder/` | [speridlabs/idol](https://huggingface.co/speridlabs/idol/tree/2731c2291fc65a797c5cb627a5e5526fc1490902) | Apache-2.0; see `licenses/Apache-2.0.txt` | |
| | `imageEncoder/` | [facebook/dinov3-vith16plus-pretrain-lvd1689m](https://huggingface.co/facebook/dinov3-vith16plus-pretrain-lvd1689m/tree/c807c9eeea853df70aec4069e6f56b28ddc82acc) | Meta DINOv3 License; see `licenses/DINOv3-LICENSE.md` | |
| | `vaeEncoder/` | [black-forest-labs/FLUX.2-dev](https://huggingface.co/black-forest-labs/FLUX.2-dev/tree/26afe3a78bb242c0a8bb181dcc8937bb16e5c66c) | Autoencoder component under Apache-2.0; see `licenses/Apache-2.0.txt` | |
|
|
| DINOv3 and its derivatives remain subject to Meta's agreement, including its |
| redistribution and use conditions. Keep that agreement with redistributed copies. |
| Black Forest Labs [licenses its FLUX.2 autoencoder under Apache-2.0](https://github.com/black-forest-labs/flux2#flux2-autoencoder). |
| Only the VAE encoder, quantization convolution and normalization statistics are |
| included; no FLUX image-generation transformer is included. |
|
|
| IDOL's reference SDK is Apache-2.0 at revision |
| `4702a7f00850e34b6e83a2170cc123c3e9570744`. IDOL acknowledges TripoSplat; |
| its MIT text and VAST copyright are retained in `licenses/TripoSplat-MIT.txt`. |
| `licenses/NOTICE` retains the reference-implementation acknowledgments. |
|
|
| Dense matrices are selectively quantized and the components repacked into stage |
| shards. Sensitive tensors and the VAE retain their converted floating-point |
| representations. No retraining was performed. These changes do not replace any |
| component's license. The separately distributed runtime's MIT license applies |
| to that runtime, not to all model weights. |
|
|
