thorjank commited on
Commit
3f62747
·
verified ·
1 Parent(s): 9792c05

Upload 6 files

Browse files
Files changed (6) hide show
  1. .job_config.json +147 -0
  2. config.yaml +135 -0
  3. latex_v1.safetensors +3 -0
  4. log.txt +0 -0
  5. optimizer.pt +3 -0
  6. pid.txt +1 -0
.job_config.json ADDED
@@ -0,0 +1,147 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "job": "extension",
3
+ "config": {
4
+ "name": "latex_v1",
5
+ "process": [
6
+ {
7
+ "type": "diffusion_trainer",
8
+ "training_folder": "/data/app/output",
9
+ "sqlite_db_path": "/data/app/aitk_db.db",
10
+ "device": "cuda",
11
+ "trigger_word": "latex",
12
+ "performance_log_every": 10,
13
+ "network": {
14
+ "type": "lora",
15
+ "linear": 32,
16
+ "linear_alpha": 32,
17
+ "conv": 16,
18
+ "conv_alpha": 16,
19
+ "lokr_full_rank": true,
20
+ "lokr_factor": -1,
21
+ "network_kwargs": {
22
+ "ignore_if_contains": []
23
+ }
24
+ },
25
+ "save": {
26
+ "dtype": "bf16",
27
+ "save_every": 250,
28
+ "max_step_saves_to_keep": 4,
29
+ "save_format": "diffusers",
30
+ "push_to_hub": false
31
+ },
32
+ "datasets": [
33
+ {
34
+ "folder_path": "/data/app/datasets/latex_01",
35
+ "mask_path": null,
36
+ "mask_min_value": 0.1,
37
+ "default_caption": "",
38
+ "caption_ext": "txt",
39
+ "caption_dropout_rate": 0.05,
40
+ "cache_latents_to_disk": false,
41
+ "is_reg": false,
42
+ "network_weight": 1,
43
+ "resolution": [
44
+ 512,
45
+ 768,
46
+ 1024
47
+ ],
48
+ "controls": [],
49
+ "shrink_video_to_frames": true,
50
+ "num_frames": 1,
51
+ "do_i2v": true,
52
+ "flip_x": false,
53
+ "flip_y": false
54
+ }
55
+ ],
56
+ "train": {
57
+ "batch_size": 1,
58
+ "bypass_guidance_embedding": false,
59
+ "steps": 3000,
60
+ "gradient_accumulation": 1,
61
+ "train_unet": true,
62
+ "train_text_encoder": false,
63
+ "gradient_checkpointing": true,
64
+ "noise_scheduler": "flowmatch",
65
+ "optimizer": "adamw8bit",
66
+ "timestep_type": "weighted",
67
+ "content_or_style": "balanced",
68
+ "optimizer_params": {
69
+ "weight_decay": 0.0001
70
+ },
71
+ "unload_text_encoder": false,
72
+ "cache_text_embeddings": false,
73
+ "lr": 0.0001,
74
+ "ema_config": {
75
+ "use_ema": false,
76
+ "ema_decay": 0.99
77
+ },
78
+ "skip_first_sample": false,
79
+ "force_first_sample": false,
80
+ "disable_sampling": false,
81
+ "dtype": "bf16",
82
+ "diff_output_preservation": false,
83
+ "diff_output_preservation_multiplier": 1,
84
+ "diff_output_preservation_class": "person",
85
+ "switch_boundary_every": 1,
86
+ "loss_type": "mse",
87
+ "do_differential_guidance": true,
88
+ "differential_guidance_scale": 3
89
+ },
90
+ "model": {
91
+ "name_or_path": "Tongyi-MAI/Z-Image-Turbo",
92
+ "quantize": true,
93
+ "qtype": "qfloat8",
94
+ "quantize_te": true,
95
+ "qtype_te": "qfloat8",
96
+ "arch": "zimage:turbo",
97
+ "low_vram": false,
98
+ "model_kwargs": {},
99
+ "layer_offloading": false,
100
+ "layer_offloading_text_encoder_percent": 1,
101
+ "layer_offloading_transformer_percent": 1,
102
+ "assistant_lora_path": "ostris/zimage_turbo_training_adapter/zimage_turbo_training_adapter_v2.safetensors"
103
+ },
104
+ "sample": {
105
+ "sampler": "flowmatch",
106
+ "sample_every": 300,
107
+ "width": 512,
108
+ "height": 512,
109
+ "samples": [
110
+ {
111
+ "prompt": "A high-detail editorial portrait of a woman wearing a sleek, shining black latex bodysuit. High neck, long sleeves. The surface shows intense specular highlights and realistic micro-creases at the joints. Soft studio lighting, 8k resolution, minimalist grey background."
112
+ },
113
+ {
114
+ "prompt": "A fashion model walking in a neon-lit city street, wearing a form-fitting red latex crop top paired with baggy denim jeans. The glossy texture of the latex contrasts with the rough fabric of the denim. Cinematic lighting, rain-slicked pavement, vibrant reflections."
115
+ },
116
+ {
117
+ "prompt": "Full body shot of a woman in a midnight blue latex catsuit performing a dynamic pose. The suit has a mirror-like finish, reflecting the surrounding environment. Sharp focus on the stretching material around the knees and elbows. Professional photography, clean composition."
118
+ },
119
+ {
120
+ "prompt": "Extreme close-up macro shot of polished purple latex fabric. Focus on the smooth, non-porous surface and the way light curves around the edges. Subtle highlights and deep shadows, showing the thickness and elasticity of the material."
121
+ },
122
+ {
123
+ "prompt": "A high-fashion avant-garde look featuring a structured emerald green latex dress with exaggerated shoulder pads. The material is highly reflective and stiff, shimmering under bright runway lights. Fashion photography, blurred audience in the background."
124
+ },
125
+ {
126
+ "prompt": "A sci-fi heroine standing in a futuristic laboratory, wearing a white latex tactical suit with silver accents. The pearlescent latex surface catches the blue and orange glow of holographic displays. Hyper-realistic, volumetric lighting, intricate details."
127
+ },
128
+ {
129
+ "prompt": "A professional fashion shot of a model in a high-gloss black latex outfit, standing next to a sleek, minimalist wall. On the wall is a high-quality acrylic sign that reads \"Valenti Dynasty\" in elegant typography. The reflective surface of the latex catches the glow from the sign. Sharp focus, cinematic studio lighting, premium aesthetic."
130
+ }
131
+ ],
132
+ "neg": "",
133
+ "seed": 42,
134
+ "walk_seed": true,
135
+ "guidance_scale": 1,
136
+ "sample_steps": 8,
137
+ "num_frames": 1,
138
+ "fps": 1
139
+ }
140
+ }
141
+ ]
142
+ },
143
+ "meta": {
144
+ "name": "[name]",
145
+ "version": "1.0"
146
+ }
147
+ }
config.yaml ADDED
@@ -0,0 +1,135 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ job: extension
2
+ config:
3
+ name: latex_v1
4
+ process:
5
+ - type: diffusion_trainer
6
+ training_folder: /data/app/output
7
+ sqlite_db_path: /data/app/aitk_db.db
8
+ device: cuda
9
+ trigger_word: latex
10
+ performance_log_every: 10
11
+ network:
12
+ type: lora
13
+ linear: 32
14
+ linear_alpha: 32
15
+ conv: 16
16
+ conv_alpha: 16
17
+ lokr_full_rank: true
18
+ lokr_factor: -1
19
+ network_kwargs:
20
+ ignore_if_contains: []
21
+ save:
22
+ dtype: bf16
23
+ save_every: 250
24
+ max_step_saves_to_keep: 4
25
+ save_format: diffusers
26
+ push_to_hub: false
27
+ datasets:
28
+ - folder_path: /data/app/datasets/latex_01
29
+ mask_path: null
30
+ mask_min_value: 0.1
31
+ default_caption: ''
32
+ caption_ext: txt
33
+ caption_dropout_rate: 0.05
34
+ cache_latents_to_disk: false
35
+ is_reg: false
36
+ network_weight: 1
37
+ resolution:
38
+ - 512
39
+ - 768
40
+ - 1024
41
+ controls: []
42
+ shrink_video_to_frames: true
43
+ num_frames: 1
44
+ do_i2v: true
45
+ flip_x: false
46
+ flip_y: false
47
+ train:
48
+ batch_size: 1
49
+ bypass_guidance_embedding: false
50
+ steps: 3000
51
+ gradient_accumulation: 1
52
+ train_unet: true
53
+ train_text_encoder: false
54
+ gradient_checkpointing: true
55
+ noise_scheduler: flowmatch
56
+ optimizer: adamw8bit
57
+ timestep_type: weighted
58
+ content_or_style: balanced
59
+ optimizer_params:
60
+ weight_decay: 0.0001
61
+ unload_text_encoder: false
62
+ cache_text_embeddings: false
63
+ lr: 0.0001
64
+ ema_config:
65
+ use_ema: false
66
+ ema_decay: 0.99
67
+ skip_first_sample: false
68
+ force_first_sample: false
69
+ disable_sampling: false
70
+ dtype: bf16
71
+ diff_output_preservation: false
72
+ diff_output_preservation_multiplier: 1
73
+ diff_output_preservation_class: person
74
+ switch_boundary_every: 1
75
+ loss_type: mse
76
+ do_differential_guidance: true
77
+ differential_guidance_scale: 3
78
+ model:
79
+ name_or_path: Tongyi-MAI/Z-Image-Turbo
80
+ quantize: true
81
+ qtype: qfloat8
82
+ quantize_te: true
83
+ qtype_te: qfloat8
84
+ arch: zimage:turbo
85
+ low_vram: false
86
+ model_kwargs: {}
87
+ layer_offloading: false
88
+ layer_offloading_text_encoder_percent: 1
89
+ layer_offloading_transformer_percent: 1
90
+ assistant_lora_path: ostris/zimage_turbo_training_adapter/zimage_turbo_training_adapter_v2.safetensors
91
+ sample:
92
+ sampler: flowmatch
93
+ sample_every: 300
94
+ width: 512
95
+ height: 512
96
+ samples:
97
+ - prompt: A high-detail editorial portrait of a woman wearing a sleek, shining
98
+ black latex bodysuit. High neck, long sleeves. The surface shows intense
99
+ specular highlights and realistic micro-creases at the joints. Soft studio
100
+ lighting, 8k resolution, minimalist grey background.
101
+ - prompt: A fashion model walking in a neon-lit city street, wearing a form-fitting
102
+ red latex crop top paired with baggy denim jeans. The glossy texture of
103
+ the latex contrasts with the rough fabric of the denim. Cinematic lighting,
104
+ rain-slicked pavement, vibrant reflections.
105
+ - prompt: Full body shot of a woman in a midnight blue latex catsuit performing
106
+ a dynamic pose. The suit has a mirror-like finish, reflecting the surrounding
107
+ environment. Sharp focus on the stretching material around the knees and
108
+ elbows. Professional photography, clean composition.
109
+ - prompt: Extreme close-up macro shot of polished purple latex fabric. Focus
110
+ on the smooth, non-porous surface and the way light curves around the edges.
111
+ Subtle highlights and deep shadows, showing the thickness and elasticity
112
+ of the material.
113
+ - prompt: A high-fashion avant-garde look featuring a structured emerald green
114
+ latex dress with exaggerated shoulder pads. The material is highly reflective
115
+ and stiff, shimmering under bright runway lights. Fashion photography, blurred
116
+ audience in the background.
117
+ - prompt: A sci-fi heroine standing in a futuristic laboratory, wearing a white
118
+ latex tactical suit with silver accents. The pearlescent latex surface catches
119
+ the blue and orange glow of holographic displays. Hyper-realistic, volumetric
120
+ lighting, intricate details.
121
+ - prompt: A professional fashion shot of a model in a high-gloss black latex
122
+ outfit, standing next to a sleek, minimalist wall. On the wall is a high-quality
123
+ acrylic sign that reads "Valenti Dynasty" in elegant typography. The reflective
124
+ surface of the latex catches the glow from the sign. Sharp focus, cinematic
125
+ studio lighting, premium aesthetic.
126
+ neg: ''
127
+ seed: 42
128
+ walk_seed: true
129
+ guidance_scale: 1
130
+ sample_steps: 8
131
+ num_frames: 1
132
+ fps: 1
133
+ meta:
134
+ name: latex_v1
135
+ version: '1.0'
latex_v1.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:facce9b84ab6bca7ca3e52aac9c47814725c1f91c545b32609a93540660ee8e1
3
+ size 170128264
log.txt ADDED
The diff for this file is too large to render. See raw diff
 
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33f0eb735c66fbb433e80a1e45ec516b43ea4174a1212f5a2a9ba9ac1df294ee
3
+ size 173288293
pid.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ 4896