Suzuki Naru commited on
Upload ho3d/train.yml with huggingface_hub
Browse files- ho3d/train.yml +119 -0
ho3d/train.yml
ADDED
|
@@ -0,0 +1,119 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
data:
|
| 2 |
+
base_dir: /work/narus/HO3D/train
|
| 3 |
+
base_path: /work/narus/HO3D/train
|
| 4 |
+
data_root: /work/narus
|
| 5 |
+
dataset: ho3d
|
| 6 |
+
dataset_path: /work/narus/HO3D/train
|
| 7 |
+
gt_joints_path: /work/narus/FreiHAND/evaluation_xyz.json
|
| 8 |
+
mixed:
|
| 9 |
+
- name: hog
|
| 10 |
+
path: /work/narus/HOGraspNet
|
| 11 |
+
weight: 1.0
|
| 12 |
+
- name: ho3d
|
| 13 |
+
path: /work/narus/HO3D/train
|
| 14 |
+
weight: 1.0
|
| 15 |
+
- name: freihand
|
| 16 |
+
path: /work/narus/FreiHAND
|
| 17 |
+
weight: 1.0
|
| 18 |
+
- name: dexycb_sampled
|
| 19 |
+
path: /work/narus/DEXYCB_sampled
|
| 20 |
+
weight: 1.0
|
| 21 |
+
num_workers: 4
|
| 22 |
+
setup: s1
|
| 23 |
+
device: cuda
|
| 24 |
+
diffusion:
|
| 25 |
+
beta_end: 0.01
|
| 26 |
+
beta_schedule: linear
|
| 27 |
+
beta_start: 0.0001
|
| 28 |
+
diffuse_space: original
|
| 29 |
+
num_diffusion_timesteps: 1000
|
| 30 |
+
model:
|
| 31 |
+
coords_dim: 45
|
| 32 |
+
default:
|
| 33 |
+
coords_dim: 45
|
| 34 |
+
custom_pe: false
|
| 35 |
+
downsample: 1
|
| 36 |
+
dropout: 0.25
|
| 37 |
+
hamer_cond_prob: 0.2
|
| 38 |
+
hid_dim: 2056
|
| 39 |
+
n_head: 4
|
| 40 |
+
num_layer: 7
|
| 41 |
+
skips:
|
| 42 |
+
- 1
|
| 43 |
+
- 3
|
| 44 |
+
- 5
|
| 45 |
+
- 6
|
| 46 |
+
unconditinal: true
|
| 47 |
+
zero_mask_token: false
|
| 48 |
+
ema: false
|
| 49 |
+
ema_rate: 0.999
|
| 50 |
+
emb_dim: 512
|
| 51 |
+
hid_dim: 2056
|
| 52 |
+
text_feature: distilbert
|
| 53 |
+
uniaff:
|
| 54 |
+
add_img_pe: true
|
| 55 |
+
coords_dim: 45
|
| 56 |
+
custom_pe: false
|
| 57 |
+
dropout: 0.25
|
| 58 |
+
hid_dim: 2056
|
| 59 |
+
mask_by_sample_img: true
|
| 60 |
+
mask_by_sample_txt: true
|
| 61 |
+
n_head: 4
|
| 62 |
+
num_layer: 1
|
| 63 |
+
zero_mask_token: false
|
| 64 |
+
var_type: fixedsmall
|
| 65 |
+
optim:
|
| 66 |
+
amsgrad: false
|
| 67 |
+
decay: 20
|
| 68 |
+
eps: 1.0e-08
|
| 69 |
+
grad_clip: 1.0
|
| 70 |
+
lr: 0.001
|
| 71 |
+
lr_gamma: 0.9
|
| 72 |
+
optimizer: Adam
|
| 73 |
+
save_samples:
|
| 74 |
+
anti_pen: true
|
| 75 |
+
caption_index: gt_taxonomy_summary_strict
|
| 76 |
+
diffusion_space: original
|
| 77 |
+
image_cond: true
|
| 78 |
+
skip: 10
|
| 79 |
+
skip_type: uniform
|
| 80 |
+
test_num_diffusion_timesteps: 1000
|
| 81 |
+
test_times: 1
|
| 82 |
+
test_timesteps: 100
|
| 83 |
+
vis: true
|
| 84 |
+
vis_batch: 2500
|
| 85 |
+
vis_epoch: 4
|
| 86 |
+
testing:
|
| 87 |
+
anti_pen: true
|
| 88 |
+
skip: 10
|
| 89 |
+
test_num_diffusion_timesteps: 1000
|
| 90 |
+
test_times: 1
|
| 91 |
+
test_timesteps: 5
|
| 92 |
+
vis: true
|
| 93 |
+
vis_batch: 2500
|
| 94 |
+
vis_epoch: 4
|
| 95 |
+
training:
|
| 96 |
+
batch_size: 512
|
| 97 |
+
custom_pe: false
|
| 98 |
+
decay: 60
|
| 99 |
+
desc_simple: true
|
| 100 |
+
hamer_cond: false
|
| 101 |
+
image_cond: true
|
| 102 |
+
image_feat_type: hamer
|
| 103 |
+
lr: 5.0e-05
|
| 104 |
+
lr_gamma: 0.9
|
| 105 |
+
mano_grad: true
|
| 106 |
+
n_epochs: 1000
|
| 107 |
+
n_iters: 5000000
|
| 108 |
+
normalize: false
|
| 109 |
+
reg_mean: true
|
| 110 |
+
reg_var: true
|
| 111 |
+
snapshot_freq: 5000
|
| 112 |
+
text_cond_drop_prob: 0.1
|
| 113 |
+
text_mask_prob: 0.1
|
| 114 |
+
unconditinal: false
|
| 115 |
+
use_all_captions: true
|
| 116 |
+
validation_freq: 2000
|
| 117 |
+
vis: true
|
| 118 |
+
vis_batch: 2500
|
| 119 |
+
vit_feat_drop_prob: 0.3
|