forked from kohya-ss/sd-scripts
-
Notifications
You must be signed in to change notification settings - Fork 4
/
config_file.toml
80 lines (71 loc) · 2.15 KB
/
config_file.toml
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
75
76
77
78
79
80
[sdxl_arguments]
cache_text_encoder_outputs = false
no_half_vae = false
min_timestep = 0
max_timestep = 1000
[model_arguments]
pretrained_model_name_or_path = "/workspace/model/animagine-xl-3.0-base.safetensors"
vae = "/workspace/vae/sdxl_vae.safetensors"
[dataset_arguments]
shuffle_caption = true
debug_dataset = false
in_json = "/workspace/fine_tune/animagine-xl-3.1_lat.json"
train_data_dir = "/workspace/train_data/animagine-xl-3.1"
dataset_repeats = 1
keep_tokens_separator = "|||"
resolution = "1024, 1024"
caption_dropout_rate = 0
caption_tag_dropout_rate = 0
caption_dropout_every_n_epochs = 0
token_warmup_min = 1
token_warmup_step = 0
[training_arguments]
output_dir = "/workspace/fine_tune/outputs/animagine-xl-3.1"
output_name = "animagine-xl-3.1"
save_precision = "fp16"
save_every_n_steps = 1000
save_last_n_steps = true
save_state = true
save_last_n_steps_state = true
train_batch_size = 16
max_token_length = 225
mem_eff_attn = false
xformers = true
sdpa = false
max_train_epochs = 10
max_data_loader_n_workers = 8
persistent_data_loader_workers = true
gradient_checkpointing = true
gradient_accumulation_steps = 3
mixed_precision = "fp16"
ddp_gradient_as_bucket_view = true
ddp_static_graph = true
ddp_timeout = 100000
[logging_arguments]
log_with = "wandb"
log_tracker_name = "animagine-xl-3.1"
logging_dir = "/workspace/fine_tune/logs"
[sample_prompt_arguments]
sample_every_n_steps = 100
sample_sampler = "euler_a"
[saving_arguments]
save_model_as = "safetensors"
[optimizer_arguments]
optimizer_type = "AdamW"
learning_rate = 1e-5
train_text_encoder = true
optimizer_args = [ "weight_decay=0.1", "betas=0.9,0.99",]
lr_scheduler = "cosine_with_restarts"
lr_scheduler_num_cycles = 10
lr_scheduler_type = "LoraEasyCustomOptimizer.CustomOptimizers.CosineAnnealingWarmupRestarts"
lr_scheduler_args = [ "min_lr=1e-06", "gamma=0.9", "first_cycle_steps=9099",]
max_grad_norm = 1.0
[advanced_training_config]
resume_from_huggingface = false
[save_to_hub_config]
huggingface_repo_type = "model"
huggingface_path_in_repo = "model/animagine-xl-3.1_20240320_104513"
huggingface_token = ""
async_upload = true
save_state_to_huggingface = true
huggingface_repo_visibility = "private"