model:

pretrained_model_name_or_path: 'pretrained_weights/TripoSG'

vae:

num_tokens: 512

transformer:

enable_local_cross_attn: true

global_attn_block_ids: [0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20]

global_attn_block_id_range: null # The average should be 10 for unet-skipping

dataset:

config:

- 'datasets/object_part_configs.json' # Modify this path if you use your own dataset

training_ratio: 0.9

min_num_parts: 1

max_num_parts: 16

max_iou_mean: 0.5

max_iou_max: 0.5

shuffle_parts: true

object_ratio: 0.3

rotating_ratio: 0.2

ratating_degree: 10

optimizer:

name: "adamw"

lr: 5e-5

betas:

- 0.9

- 0.999

weight_decay: 0.01

eps: 1.e-8

lr_scheduler:

name: "constant_warmup"

num_warmup_steps: 1000

train:

batch_size_per_gpu: 32

epochs: 10

grad_checkpoint: true

weighting_scheme: "logit_normal"

logit_mean: 0.0

logit_std: 1.0

mode_scale: 1.29

cfg_dropout_prob: 0.1

training_objective: "-v"

log_freq: 1

early_eval_freq: 500

early_eval: 1000

eval_freq: 1000

save_freq: 2000

eval_freq_epoch: 5

save_freq_epoch: 10

ema_kwargs:

decay: 0.9999

use_ema_warmup: true

inv_gamma: 1.

power: 0.75

val:

batch_size_per_gpu: 1

nrow: 4

min_num_parts: 2

max_num_parts: 8

num_inference_steps: 50

max_num_expanded_coords: 1e8

use_flash_decoder: false

rendering:

radius: 4.0

num_views: 36

fps: 18

metric:

cd_num_samples: 204800

cd_metric: "l2"

f1_score_threshold: 0.1

default_cd: 1e6

default_f1: 0.0