File size: 1,538 Bytes
559d631
 
 
 
 
 
 
8679f41
559d631
 
 
 
 
 
 
 
8679f41
559d631
8679f41
 
559d631
 
8679f41
 
 
559d631
 
 
 
 
 
 
8679f41
559d631
 
 
 
 
 
8679f41
 
559d631
 
 
 
8679f41
 
559d631
 
 
8679f41
 
559d631
 
 
 
 
 
 
 
 
 
 
 
 
8679f41
559d631
 
 
 
 
8679f41
559d631
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1.0e-08
adam_weight_decay: 0.01
algorithm_type: dpmsolver++
beta_scheduler: scaled_linear
channel_attn: true
ckpt_interval: 1
content_encoder_downsample_size: 3
content_image_size: !!python/tuple
- 96
- 96
content_start_channel: 64
correcting_x0_fn: null
data_root: my_dataset
drop_prob: 0.1
enable_style_transform: false
experience_name: FontDiffuser_training_phase_2
export_onnx_path: null
gradient_accumulation_steps: 1
guidance_scale: 7.5
guidance_type: classifier-free
hf_repo_id: dzungpham/font-diffusion-generated-data
hf_split: handwritten_original
hf_token: null
learning_rate: 1.0e-05
local_rank: -1
log_interval: 50
logging_dir: logs
lr_scheduler: constant
lr_warmup_steps: 100
max_grad_norm: 1.0
max_train_steps: 2
method: multistep
mixed_precision: 'no'
mode: refinement
model_type: noise
nce_layers: 0,1,2,3
num_inference_steps: 20
num_neg: 12
num_workers: 4
offset_coefficient: 0.5
order: 2
output_dir: outputs/FontDiffuser
perceptual_coefficient: 0.05
phase_1: false
phase_1_ckpt_dir: outputs/FontDiffuser/checkpoint_step_1000
phase_2: true
report_to: wandb
resolution: 96
resume_from_checkpoint: null
save_full_model: false
sc_coefficient: 0.05
scale_lr: false
scr_ckpt_path: ckpt/scr_210000.pth
scr_image_size: 96
seed: 123
skip_type: time_uniform
style_image_size: !!python/tuple
- 96
- 96
style_start_channel: 64
t_end: null
t_start: null
temperature: 0.07
train_batch_size: 16
unet_channels: !!python/tuple
- 64
- 128
- 256
- 512
use_hf_dataset: true
val_interval: 100