| model: | |
| base_learning_rate: 3.0e-05 | |
| target: svrm.ldm.models.svrm.SVRMModel | |
| params: | |
| img_encoder_config: | |
| target: svrm.ldm.modules.encoders.dinov2_mod.FrozenDinoV2ImageEmbedder | |
| params: | |
| version: dinov2_vitb14 | |
| img_to_triplane_config: | |
| target: svrm.ldm.modules.translator.img_to_triplane.ImgToTriplaneModel | |
| params: | |
| pos_emb_size: 64 | |
| pos_emb_dim: 1024 | |
| cam_cond_dim: 20 | |
| n_heads: 16 | |
| d_head: 64 | |
| depth: 16 | |
| context_dim: 768 | |
| triplane_dim: 120 | |
| use_fp16: true | |
| use_bf16: false | |
| upsample_time: 2 | |
| render_config: | |
| target: svrm.ldm.modules.rendering_neus.synthesizer.TriplaneSynthesizer | |
| params: | |
| triplane_dim: 120 | |
| samples_per_ray: 128 | |