spawn99 commited on
Commit
98f5620
1 Parent(s): a5758a0

Model card auto-generated by SimpleTuner

Browse files
Files changed (1) hide show
  1. README.md +142 -0
README.md ADDED
@@ -0,0 +1,142 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: other
3
+ base_model: "black-forest-labs/FLUX.1-dev"
4
+ tags:
5
+ - flux
6
+ - flux-diffusers
7
+ - text-to-image
8
+ - diffusers
9
+ - simpletuner
10
+ - safe-for-work
11
+ - lora
12
+ - template:sd-lora
13
+ - standard
14
+ inference: true
15
+ widget:
16
+ - text: 'unconditional (blank prompt)'
17
+ parameters:
18
+ negative_prompt: 'blurry, cropped, ugly'
19
+ output:
20
+ url: ./assets/image_0_0.png
21
+ - text: 'photo of a modern architectural interior event space at dusk'
22
+ parameters:
23
+ negative_prompt: 'blurry, cropped, ugly'
24
+ output:
25
+ url: ./assets/image_1_0.png
26
+ ---
27
+
28
+ # simpletuner-lora-spindrift
29
+
30
+ This is a standard PEFT LoRA derived from [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev).
31
+
32
+
33
+ The main validation prompt used during training was:
34
+
35
+
36
+
37
+ ```
38
+ photo of a modern architectural interior event space at dusk
39
+ ```
40
+
41
+ ## Validation settings
42
+ - CFG: `3.0`
43
+ - CFG Rescale: `0.0`
44
+ - Steps: `25`
45
+ - Sampler: `None`
46
+ - Seed: `42`
47
+ - Resolution: `1216x832`
48
+
49
+ Note: The validation settings are not necessarily the same as the [training settings](#training-settings).
50
+
51
+ You can find some example images in the following gallery:
52
+
53
+
54
+ <Gallery />
55
+
56
+ The text encoder **was not** trained.
57
+ You may reuse the base model text encoder for inference.
58
+
59
+
60
+ ## Training settings
61
+
62
+ - Training epochs: 0
63
+ - Training steps: 1000
64
+ - Learning rate: 1e-05
65
+ - Effective batch size: 1
66
+ - Micro-batch size: 1
67
+ - Gradient accumulation steps: 1
68
+ - Number of GPUs: 1
69
+ - Prediction type: flow-matching
70
+ - Rescaled betas zero SNR: False
71
+ - Optimizer: adamw_bf16
72
+ - Precision: Pure BF16
73
+ - Quantised: Yes: int8-quanto
74
+ - Xformers: Not used
75
+ - LoRA Rank: 64
76
+ - LoRA Alpha: None
77
+ - LoRA Dropout: 0.1
78
+ - LoRA initialisation style: default
79
+
80
+
81
+ ## Datasets
82
+
83
+ ### spindrift-dataset-512
84
+ - Repeats: 25
85
+ - Total number of images: 29
86
+ - Total number of aspect buckets: 6
87
+ - Resolution: 0.262144 megapixels
88
+ - Cropped: False
89
+ - Crop style: None
90
+ - Crop aspect: None
91
+ ### spindrift-dataset-1024
92
+ - Repeats: 25
93
+ - Total number of images: 32
94
+ - Total number of aspect buckets: 1
95
+ - Resolution: 1.048576 megapixels
96
+ - Cropped: False
97
+ - Crop style: None
98
+ - Crop aspect: None
99
+ ### spindrift-dataset-512-crop
100
+ - Repeats: 25
101
+ - Total number of images: 32
102
+ - Total number of aspect buckets: 1
103
+ - Resolution: 0.262144 megapixels
104
+ - Cropped: True
105
+ - Crop style: random
106
+ - Crop aspect: square
107
+ ### spindrift-dataset-1024-crop
108
+ - Repeats: 25
109
+ - Total number of images: 32
110
+ - Total number of aspect buckets: 1
111
+ - Resolution: 1.048576 megapixels
112
+ - Cropped: True
113
+ - Crop style: random
114
+ - Crop aspect: square
115
+
116
+
117
+ ## Inference
118
+
119
+
120
+ ```python
121
+ import torch
122
+ from diffusers import DiffusionPipeline
123
+
124
+ model_id = 'black-forest-labs/FLUX.1-dev'
125
+ adapter_id = 'spawn99/simpletuner-lora-spindrift'
126
+ pipeline = DiffusionPipeline.from_pretrained(model_id)
127
+ pipeline.load_lora_weights(adapter_id)
128
+
129
+ prompt = "photo of a modern architectural interior event space at dusk"
130
+
131
+ pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
132
+ image = pipeline(
133
+ prompt=prompt,
134
+ num_inference_steps=25,
135
+ generator=torch.Generator(device='cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu').manual_seed(1641421826),
136
+ width=1216,
137
+ height=832,
138
+ guidance_scale=3.0,
139
+ ).images[0]
140
+ image.save("output.png", format="PNG")
141
+ ```
142
+