Upload 2 files
Browse files- README.md +27 -3
- config.json +50 -0
README.md
CHANGED
|
@@ -1,3 +1,27 @@
|
|
| 1 |
-
---
|
| 2 |
-
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
tags:
|
| 3 |
+
- model_hub_mixin
|
| 4 |
+
- pytorch_model_hub_mixin
|
| 5 |
+
- renderformer
|
| 6 |
+
license: mit
|
| 7 |
+
---
|
| 8 |
+
|
| 9 |
+
# RenderFormer: Transformer-based Neural Rendering of Triangle Meshes with Global Illumination</h1>
|
| 10 |
+
|
| 11 |
+
This repo contains the weights of **RenderFormer-V1.1-Large**.
|
| 12 |
+
|
| 13 |
+
## Quick Start
|
| 14 |
+
|
| 15 |
+
Please refer to our [Github Repo](https://github.com/microsoft/renderformer).
|
| 16 |
+
|
| 17 |
+
## Citation
|
| 18 |
+
If you find our repository useful, please cite our paper in your work:
|
| 19 |
+
|
| 20 |
+
```bibtex
|
| 21 |
+
@inproceedings {zeng2025renderformer,
|
| 22 |
+
title = {RenderFormer: Transformer-based Neural Rendering of Triangle Meshes with Global Illumination},
|
| 23 |
+
author = {Chong Zeng and Yue Dong and Pieter Peers and Hongzhi Wu and Xin Tong},
|
| 24 |
+
booktitle = {ACM SIGGRAPH 2025 Conference Papers},
|
| 25 |
+
year = {2025}
|
| 26 |
+
}
|
| 27 |
+
```
|
config.json
ADDED
|
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"activation": "swiglu",
|
| 3 |
+
"bias": false,
|
| 4 |
+
"dim_feedforward": 4096,
|
| 5 |
+
"dpt_features": 256,
|
| 6 |
+
"dpt_out_channels": [
|
| 7 |
+
256,
|
| 8 |
+
512,
|
| 9 |
+
1024,
|
| 10 |
+
1024
|
| 11 |
+
],
|
| 12 |
+
"dpt_out_layers": [
|
| 13 |
+
2,
|
| 14 |
+
5,
|
| 15 |
+
8,
|
| 16 |
+
11
|
| 17 |
+
],
|
| 18 |
+
"dropout": 0.0,
|
| 19 |
+
"include_alpha": false,
|
| 20 |
+
"latent_dim": 1024,
|
| 21 |
+
"norm_first": true,
|
| 22 |
+
"norm_type": "rms_norm",
|
| 23 |
+
"num_heads": 8,
|
| 24 |
+
"num_layers": 12,
|
| 25 |
+
"num_register_tokens": 32,
|
| 26 |
+
"patch_size": 8,
|
| 27 |
+
"pe_type": "rope",
|
| 28 |
+
"qk_norm": true,
|
| 29 |
+
"rope_double_max_freq": true,
|
| 30 |
+
"rope_type": "triangle",
|
| 31 |
+
"texture_channels": 13,
|
| 32 |
+
"texture_encode_patch_size": 1,
|
| 33 |
+
"texture_encoder_norm_type": "rms_norm",
|
| 34 |
+
"turn_to_cam_coord": true,
|
| 35 |
+
"use_dpt_decoder": true,
|
| 36 |
+
"use_ldr": false,
|
| 37 |
+
"use_vn_encoder": true,
|
| 38 |
+
"vdir_num_freqs": 0,
|
| 39 |
+
"vdir_pe_type": "nerf",
|
| 40 |
+
"vertex_pe_num_freqs": 12,
|
| 41 |
+
"view_indep_qk_norm": true,
|
| 42 |
+
"view_transformer_ffn_hidden_dim": 4096,
|
| 43 |
+
"view_transformer_include_self_attn": true,
|
| 44 |
+
"view_transformer_latent_dim": 1024,
|
| 45 |
+
"view_transformer_n_heads": 8,
|
| 46 |
+
"view_transformer_n_layers": 12,
|
| 47 |
+
"view_transformer_use_swin_attn": true,
|
| 48 |
+
"vn_encoder_norm_type": "rms_norm",
|
| 49 |
+
"vn_pe_num_freqs": 6
|
| 50 |
+
}
|