Naznut commited on
Commit
b1b5245
·
verified ·
1 Parent(s): ebfee0f

Upload folder using huggingface_hub

Browse files
Files changed (3) hide show
  1. README.md +9 -2
  2. config.yaml +30 -7
  3. lora.safetensors +1 -1
README.md CHANGED
@@ -11,18 +11,25 @@ tags:
11
  - replicate
12
  base_model: "black-forest-labs/FLUX.1-dev"
13
  pipeline_tag: text-to-image
14
- instance_prompt: ohwx
 
 
 
 
 
15
  ---
16
 
17
  # Pony_Loras
18
 
 
 
19
  Trained on Replicate using:
20
 
21
  https://replicate.com/ostris/flux-dev-lora-trainer/train
22
 
23
 
24
  ## Trigger words
25
- You should use `ohwx` to trigger the image generation.
26
 
27
 
28
  ## Use it with the [🧨 diffusers library](https://github.com/huggingface/diffusers)
 
11
  - replicate
12
  base_model: "black-forest-labs/FLUX.1-dev"
13
  pipeline_tag: text-to-image
14
+ # widget:
15
+ # - text: >-
16
+ # prompt
17
+ # output:
18
+ # url: https://...
19
+ instance_prompt: KRHN
20
  ---
21
 
22
  # Pony_Loras
23
 
24
+ <!-- <Gallery /> -->
25
+
26
  Trained on Replicate using:
27
 
28
  https://replicate.com/ostris/flux-dev-lora-trainer/train
29
 
30
 
31
  ## Trigger words
32
+ You should use `KRHN` to trigger the image generation.
33
 
34
 
35
  ## Use it with the [🧨 diffusers library](https://github.com/huggingface/diffusers)
config.yaml CHANGED
@@ -5,21 +5,22 @@ config:
5
  - type: custom_sd_trainer
6
  training_folder: output
7
  device: cuda:0
8
- trigger_word: ohwx
9
  network:
10
  type: lora
11
  linear: 16
12
  linear_alpha: 16
13
  save:
14
  dtype: float16
15
- save_every: 2001
16
  max_step_saves_to_keep: 1
17
  datasets:
18
  - folder_path: input_images
19
  caption_ext: txt
20
  caption_dropout_rate: 0.05
21
  shuffle_tokens: false
22
- cache_latents_to_disk: true
 
23
  resolution:
24
  - 512
25
  - 768
@@ -45,15 +46,37 @@ config:
45
  quantize: true
46
  sample:
47
  sampler: flowmatch
48
- sample_every: 2001
49
  width: 1024
50
  height: 1024
51
- prompts: []
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
52
  neg: ''
53
  seed: 42
54
  walk_seed: true
55
- guidance_scale: 4
56
- sample_steps: 20
57
  meta:
58
  name: flux_train_replicate
59
  version: '1.0'
 
5
  - type: custom_sd_trainer
6
  training_folder: output
7
  device: cuda:0
8
+ trigger_word: KRHN
9
  network:
10
  type: lora
11
  linear: 16
12
  linear_alpha: 16
13
  save:
14
  dtype: float16
15
+ save_every: 300
16
  max_step_saves_to_keep: 1
17
  datasets:
18
  - folder_path: input_images
19
  caption_ext: txt
20
  caption_dropout_rate: 0.05
21
  shuffle_tokens: false
22
+ cache_latents_to_disk: false
23
+ cache_latents: true
24
  resolution:
25
  - 512
26
  - 768
 
46
  quantize: true
47
  sample:
48
  sampler: flowmatch
49
+ sample_every: 300
50
  width: 1024
51
  height: 1024
52
+ prompts:
53
+ - a photo of KRHN, capture a mystical, high-contrast photograph of a stealthy
54
+ ninja woman posing elegantly within the intricately carved, lantern-lit chambers
55
+ of a centuries-old chinese temple. she stands poised, one hand grasping the
56
+ hilt of her katana, its scabbard adorned with crimson tassels, while her piercing
57
+ orange eyes seem to hold secrets of the ancient arts. the warm, golden light
58
+ of lanterns dancing across the walls casts dramatic shadows, illuminating
59
+ the ornate tapestries and forgotten relics scattered throughout the chamber,
60
+ transporting the viewer to a realm of mystique and intrigue.
61
+ - 'a photo of KRHN, capture a stunning, high-contrast photograph of a stealthy
62
+ female ninja standing majestically on the lunar surface, illuminated by the
63
+ radiant earthshine. her sleek black ninja garb blends seamlessly with the
64
+ shadows, except for the subtle, crescent moon emblem on her chest, which glows
65
+ softly like a beacon. her piercing orange eyes seem to pierce through the
66
+ darkness, fixed intently on some distant target. in the background, the curvature
67
+ of the earth rises like a glowing blue horizon, surrounded by the endless
68
+ expanse of starry space, with the milky way streaking across the sky like
69
+ a celestial banner. the ninja''s stance is poised and ready, as if prepared
70
+ to leap into action at a moment''s notice, her slender figure silhouetted
71
+ against the vibrant, dreamlike backdrop of our home planet. composition: low-angle
72
+ shot, looking up at the ninja from the lunar surface, with the earth and stars
73
+ stretching out above her.'
74
+ - a photo of KRHN
75
  neg: ''
76
  seed: 42
77
  walk_seed: true
78
+ guidance_scale: 3.5
79
+ sample_steps: 28
80
  meta:
81
  name: flux_train_replicate
82
  version: '1.0'
lora.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8bea14b9c234940e1ed535f5eff19382e4e2cefe8c504f680db35729ac69397c
3
  size 171969416
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a353cd3f5cafe1bc0e3abbf59d842fecd2092c5f6222d26e3116b001a86bb1a5
3
  size 171969416