Commit
•
32a1cfe
1
Parent(s):
b9572d7
Upload folder using huggingface_hub
Browse files- README.md +38 -3
- config.yaml +59 -0
- lora.safetensors +3 -0
README.md
CHANGED
@@ -1,3 +1,38 @@
|
|
1 |
-
---
|
2 |
-
license:
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: other
|
3 |
+
license_name: flux-1-dev-non-commercial-license
|
4 |
+
license_link: https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md
|
5 |
+
language:
|
6 |
+
- en
|
7 |
+
tags:
|
8 |
+
- flux
|
9 |
+
- diffusers
|
10 |
+
- lora
|
11 |
+
base_model: "black-forest-labs/FLUX.1-dev"
|
12 |
+
pipeline_tag: text-to-image
|
13 |
+
instance_prompt: PURLPICS
|
14 |
+
---
|
15 |
+
|
16 |
+
# Purlpics Lora
|
17 |
+
|
18 |
+
Trained on Replicate using:
|
19 |
+
|
20 |
+
https://replicate.com/ostris/flux-dev-lora-trainer/train
|
21 |
+
|
22 |
+
|
23 |
+
## Trigger words
|
24 |
+
You should use `PURLPICS` to trigger the image generation.
|
25 |
+
|
26 |
+
|
27 |
+
## Use it with the [🧨 diffusers library](https://github.com/huggingface/diffusers)
|
28 |
+
|
29 |
+
```py
|
30 |
+
from diffusers import AutoPipelineForText2Image
|
31 |
+
import torch
|
32 |
+
|
33 |
+
pipeline = AutoPipelineForText2Image.from_pretrained('black-forest-labs/FLUX.1-dev', torch_dtype=torch.float16).to('cuda')
|
34 |
+
pipeline.load_lora_weights('levindixon/purlpics-lora', weight_name='lora.safetensors')
|
35 |
+
image = pipeline('your prompt').images[0]
|
36 |
+
```
|
37 |
+
|
38 |
+
For more details, including weighting, merging and fusing LoRAs, check the [documentation on loading LoRAs in diffusers](https://huggingface.co/docs/diffusers/main/en/using-diffusers/loading_adapters)
|
config.yaml
ADDED
@@ -0,0 +1,59 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
job: custom_job
|
2 |
+
config:
|
3 |
+
name: flux_train_replicate
|
4 |
+
process:
|
5 |
+
- type: custom_sd_trainer
|
6 |
+
training_folder: output
|
7 |
+
device: cuda:0
|
8 |
+
trigger_word: PURLPICS
|
9 |
+
network:
|
10 |
+
type: lora
|
11 |
+
linear: 64
|
12 |
+
linear_alpha: 64
|
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
|
26 |
+
- 1024
|
27 |
+
train:
|
28 |
+
batch_size: 1
|
29 |
+
steps: 2000
|
30 |
+
gradient_accumulation_steps: 1
|
31 |
+
train_unet: true
|
32 |
+
train_text_encoder: false
|
33 |
+
content_or_style: balanced
|
34 |
+
gradient_checkpointing: true
|
35 |
+
noise_scheduler: flowmatch
|
36 |
+
optimizer: adamw8bit
|
37 |
+
lr: 0.0004
|
38 |
+
ema_config:
|
39 |
+
use_ema: true
|
40 |
+
ema_decay: 0.99
|
41 |
+
dtype: bf16
|
42 |
+
model:
|
43 |
+
name_or_path: FLUX.1-dev
|
44 |
+
is_flux: true
|
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'
|
lora.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:df0ded967258622d12627b6827b637e4d9bb23f73056eee05509b03d937c70cf
|
3 |
+
size 687476576
|