pytorch3d/projects/implicitron_trainer/configs/repro_singleseq_srn.yaml
David Novotny e767c4b548 Raysampler as pluggable
Summary:
This converts raysamplers to ReplaceableBase so that users can hack their own raysampling impls.

Context: Andrea tried to implement TensoRF within implicitron but could not due to the need to implement his own raysampler.

Reviewed By: shapovalov

Differential Revision: D36016318

fbshipit-source-id: ef746f3365282bdfa9c15f7b371090a5aae7f8da
2022-05-12 15:39:35 -07:00

29 lines
779 B
YAML

defaults:
- repro_singleseq_base.yaml
- _self_
generic_model_args:
num_passes: 1
chunk_size_grid: 32000
view_pooler_enabled: false
loss_weights:
loss_rgb_mse: 200.0
loss_prev_stage_rgb_mse: 0.0
loss_mask_bce: 1.0
loss_prev_stage_mask_bce: 0.0
loss_autodecoder_norm: 0.0
depth_neg_penalty: 10000.0
raysampler_class_type: NearFarRaySampler
raysampler_NearFarRaySampler_args:
n_rays_per_image_sampled_from_mask: 2048
min_depth: 0.05
max_depth: 0.05
n_pts_per_ray_training: 1
n_pts_per_ray_evaluation: 1
stratified_point_sampling_training: false
stratified_point_sampling_evaluation: false
renderer_class_type: LSTMRenderer
implicit_function_class_type: SRNImplicitFunction
solver_args:
breed: adam
lr: 5.0e-05