Delete ckpts/image-to-shape-diffusion
Browse files
ckpts/image-to-shape-diffusion/clip-mvrgb-modln-l256-e64-ne8-nd16-nl6/config.yaml
DELETED
|
@@ -1,143 +0,0 @@
|
|
| 1 |
-
name: michelangelo-image-to-shape-diffusion/clip-mvrgb-modln-l256-e64-ne8-nd16-nl6-170k
|
| 2 |
-
description: ''
|
| 3 |
-
tag: michelangelo-aligned-autoencoder+n4096+noise0.0+pfeat3+zeroemb0.0+normembFalse+lr5e-05+qkvbiasFalse+nfreq8+ln_postTrue
|
| 4 |
-
seed: 0
|
| 5 |
-
use_timestamp: true
|
| 6 |
-
timestamp: ''
|
| 7 |
-
exp_root_dir: outputs
|
| 8 |
-
exp_dir: outputs/michelangelo-image-to-shape-diffusion/clip-mvrgb-modln-l256-e64-ne8-nd16-nl6-170k
|
| 9 |
-
trial_name: michelangelo-aligned-autoencoder+n4096+noise0.0+pfeat3+zeroemb0.0+normembFalse+lr5e-05+qkvbiasFalse+nfreq8+ln_postTrue
|
| 10 |
-
trial_dir: outputs/michelangelo-image-to-shape-diffusion/clip-mvrgb-modln-l256-e64-ne8-nd16-nl6-170k/michelangelo-aligned-autoencoder+n4096+noise0.0+pfeat3+zeroemb0.0+normembFalse+lr5e-05+qkvbiasFalse+nfreq8+ln_postTrue
|
| 11 |
-
n_gpus: 8
|
| 12 |
-
resume: ./ckpts/3DNativeGeneration/michelangelo-image-to-shape-diffusion/clip-mvrgb-modln-l256-e64-ne8-nd16-nl6-170k.ckpt
|
| 13 |
-
data_type: objaverse-datamodule
|
| 14 |
-
data:
|
| 15 |
-
root_dir: data/objaverse_clean/cap3d_high_quality_170k_images
|
| 16 |
-
data_type: occupancy
|
| 17 |
-
n_samples: 4096
|
| 18 |
-
noise_sigma: 0.0
|
| 19 |
-
load_supervision: false
|
| 20 |
-
supervision_type: occupancy
|
| 21 |
-
n_supervision: 10000
|
| 22 |
-
load_image: true
|
| 23 |
-
image_data_path: data/objaverse_clean/raw_data/images/cap3d_high_quality_170k
|
| 24 |
-
image_type: mvrgb
|
| 25 |
-
idx:
|
| 26 |
-
- 0
|
| 27 |
-
- 4
|
| 28 |
-
- 8
|
| 29 |
-
- 12
|
| 30 |
-
- 16
|
| 31 |
-
n_views: 4
|
| 32 |
-
load_caption: false
|
| 33 |
-
rotate_points: false
|
| 34 |
-
batch_size: 32
|
| 35 |
-
num_workers: 16
|
| 36 |
-
system_type: shape-diffusion-system
|
| 37 |
-
system:
|
| 38 |
-
val_samples_json: val_data/mv_images/val_samples_rgb_mvimage.json
|
| 39 |
-
z_scale_factor: 1.0
|
| 40 |
-
guidance_scale: 7.5
|
| 41 |
-
num_inference_steps: 50
|
| 42 |
-
eta: 0.0
|
| 43 |
-
shape_model_type: michelangelo-aligned-autoencoder
|
| 44 |
-
shape_model:
|
| 45 |
-
num_latents: 256
|
| 46 |
-
embed_dim: 64
|
| 47 |
-
point_feats: 3
|
| 48 |
-
out_dim: 1
|
| 49 |
-
num_freqs: 8
|
| 50 |
-
include_pi: false
|
| 51 |
-
heads: 12
|
| 52 |
-
width: 768
|
| 53 |
-
num_encoder_layers: 8
|
| 54 |
-
num_decoder_layers: 16
|
| 55 |
-
use_ln_post: true
|
| 56 |
-
init_scale: 0.25
|
| 57 |
-
qkv_bias: false
|
| 58 |
-
use_flash: true
|
| 59 |
-
use_checkpoint: true
|
| 60 |
-
condition_model_type: clip-embedder
|
| 61 |
-
condition_model:
|
| 62 |
-
pretrained_model_name_or_path: openai/clip-vit-large-patch14
|
| 63 |
-
encode_camera: true
|
| 64 |
-
camera_embeds_dim: 32
|
| 65 |
-
n_views: 4
|
| 66 |
-
empty_embeds_ratio: 0.1
|
| 67 |
-
normalize_embeds: false
|
| 68 |
-
zero_uncond_embeds: true
|
| 69 |
-
denoiser_model_type: simple-denoiser
|
| 70 |
-
denoiser_model:
|
| 71 |
-
input_channels: 64
|
| 72 |
-
output_channels: 64
|
| 73 |
-
n_ctx: 256
|
| 74 |
-
width: 768
|
| 75 |
-
layers: 6
|
| 76 |
-
heads: 12
|
| 77 |
-
context_dim: 1024
|
| 78 |
-
init_scale: 1.0
|
| 79 |
-
skip_ln: true
|
| 80 |
-
use_checkpoint: true
|
| 81 |
-
noise_scheduler_type: diffusers.schedulers.DDPMScheduler
|
| 82 |
-
noise_scheduler:
|
| 83 |
-
num_train_timesteps: 1000
|
| 84 |
-
beta_start: 0.00085
|
| 85 |
-
beta_end: 0.012
|
| 86 |
-
beta_schedule: scaled_linear
|
| 87 |
-
variance_type: fixed_small
|
| 88 |
-
clip_sample: false
|
| 89 |
-
denoise_scheduler_type: diffusers.schedulers.DDIMScheduler
|
| 90 |
-
denoise_scheduler:
|
| 91 |
-
num_train_timesteps: 1000
|
| 92 |
-
beta_start: 0.00085
|
| 93 |
-
beta_end: 0.012
|
| 94 |
-
beta_schedule: scaled_linear
|
| 95 |
-
clip_sample: false
|
| 96 |
-
set_alpha_to_one: false
|
| 97 |
-
steps_offset: 1
|
| 98 |
-
loggers:
|
| 99 |
-
wandb:
|
| 100 |
-
enable: false
|
| 101 |
-
project: JiangXin
|
| 102 |
-
name: text-to-shape-diffusion+michelangelo-image-to-shape-diffusion/clip-mvrgb-modln-l256-e64-ne8-nd16-nl6-170k+michelangelo-aligned-autoencoder+n4096+noise0.0+pfeat3+zeroemb0.0+normembFalse+lr5e-05+qkvbiasFalse+nfreq8+ln_postTrue
|
| 103 |
-
loss:
|
| 104 |
-
loss_type: mse
|
| 105 |
-
lambda_diffusion: 1.0
|
| 106 |
-
optimizer:
|
| 107 |
-
name: AdamW
|
| 108 |
-
args:
|
| 109 |
-
lr: 5.0e-05
|
| 110 |
-
betas:
|
| 111 |
-
- 0.9
|
| 112 |
-
- 0.99
|
| 113 |
-
eps: 1.0e-06
|
| 114 |
-
scheduler:
|
| 115 |
-
name: SequentialLR
|
| 116 |
-
interval: step
|
| 117 |
-
schedulers:
|
| 118 |
-
- name: LinearLR
|
| 119 |
-
interval: step
|
| 120 |
-
args:
|
| 121 |
-
start_factor: 1.0e-06
|
| 122 |
-
end_factor: 1.0
|
| 123 |
-
total_iters: 5000
|
| 124 |
-
- name: CosineAnnealingLR
|
| 125 |
-
interval: step
|
| 126 |
-
args:
|
| 127 |
-
T_max: 5000
|
| 128 |
-
eta_min: 0.0
|
| 129 |
-
milestones:
|
| 130 |
-
- 5000
|
| 131 |
-
trainer:
|
| 132 |
-
num_nodes: 2
|
| 133 |
-
max_epochs: 100000
|
| 134 |
-
log_every_n_steps: 5
|
| 135 |
-
num_sanity_val_steps: 1
|
| 136 |
-
check_val_every_n_epoch: 3
|
| 137 |
-
enable_progress_bar: true
|
| 138 |
-
precision: 16-mixed
|
| 139 |
-
strategy: ddp_find_unused_parameters_true
|
| 140 |
-
checkpoint:
|
| 141 |
-
save_last: true
|
| 142 |
-
save_top_k: -1
|
| 143 |
-
every_n_train_steps: 5000
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ckpts/image-to-shape-diffusion/clip-mvrgb-modln-l256-e64-ne8-nd16-nl6/model.ckpt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:41248dba953cad356c491e7584b4171920f2ad95af10b0f78225eda867dbb7c4
|
| 3 |
-
size 3722911570
|
|
|
|
|
|
|
|
|
|
|
|