Image-to-Image
GaussianFormer
Safetensors
3d-gaussian-splatting
neural-rendering
novel-view-synthesis
transformer
pytorch_model_hub_mixin
model_hub_mixin
Instructions to use shahafvl/gaussianformer with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- GaussianFormer
How to use shahafvl/gaussianformer with GaussianFormer:
# pip install "gaussianformer @ git+https://github.com/SVLwoof/gaussianformer" from gaussianformer import GaussianFormerRenderingPipeline, load_ply pipeline = GaussianFormerRenderingPipeline.from_pretrained("shahafvl/gaussianformer").to("cuda") gaussians = load_ply("object.ply") # 3D Gaussian Splatting PLY -> [N, 14] - Notebooks
- Google Colab
- Kaggle
Download config.json from shahafvl/gaussianformer: direct link, hf CLI and curl.
- Browser
- Download file 618 Bytes
-
https://huggingface.co/shahafvl/gaussianformer/resolve/main/config.json
- Command line
-
hf download hf://shahafvl/gaussianformer/config.json
-
curl -L -o config.json https://huggingface.co/shahafvl/gaussianformer/resolve/main/config.json
618 Bytes
| { | |
| "latent_dim": 768, | |
| "num_layers": 12, | |
| "num_heads": 6, | |
| "dim_feedforward": 3072, | |
| "num_register_tokens": 16, | |
| "pos_pe_num_freqs": 12, | |
| "view_transformer_latent_dim": 768, | |
| "view_transformer_ffn_hidden_dim": 3072, | |
| "view_transformer_n_heads": 6, | |
| "view_transformer_n_layers": 6, | |
| "patch_size": 4, | |
| "ray_embed_patch": 8, | |
| "dpt_features": 128, | |
| "dpt_out_channels": [ | |
| 96, | |
| 192, | |
| 384, | |
| 768 | |
| ], | |
| "proj_rope_2d": true, | |
| "ray_rope_2d_dim": 16, | |
| "ray_rope_2d_scale": 0.25, | |
| "xattn_window": 8, | |
| "xattn_margin": 1.0, | |
| "xattn_sigmas": 3.0, | |
| "view_bf16": true, | |
| "view_grad_checkpoint": true | |
| } |