Text-to-Image
Diffusers
template:sd-lora
flux
lora
ddickinson_dstyl3xl / config.yaml
rvorias's picture
Upload folder using huggingface_hub
e3d5496 verified
config:
name: dstyl3xl
process:
- datasets:
- cache_latents_to_disk: true
caption_dropout_rate: 0.2
caption_ext: txt
folder_path: /root/lorahub/dstyl3xl/dataset
resolution:
- 512
- 768
- 1024
shuffle_tokens: false
token_dropout_rate: 0.01
device: cuda:0
model:
is_flux: true
name_or_path: black-forest-labs/FLUX.1-dev
quantize: true
text_encoder_bits: 8
network:
linear: 42
linear_alpha: 42
transformer_only: true
type: lora
performance_log_every: 500
sample:
height: 1024
neg: ''
prompts:
- Pilot flying a hot air balloon shaped like a teapot, pouring rain on dry lands
below [trigger]
- A gigantic clockwork heart powering a mechanical forest, with gears and springs
as leaves and branches [trigger]
- Scientist with wild hair examining a glowing crystal, bioluminescent plants
in the laboratory [trigger]
- A time-traveling DeLorean car emerging from a swirling portal in the middle
of a medieval village [trigger]
- A group of friendly aliens having a picnic on the moon, Earth visible in the
background [trigger]
- A Victorian-era robot having a tea party with stuffed animals in a sunlit
garden [trigger]
sample_every: 500
sample_steps: 25
sampler: flowmatch
seed: 42
walk_seed: true
width: 1024
save:
dtype: float16
max_step_saves_to_keep: 3
save_every: 500
save_format: diffusers
train:
batch_size: 1
dtype: bf16
ema_config:
ema_decay: 0.99
use_ema: true
gradient_accumulation_steps: 1
gradient_checkpointing: true
linear_timesteps: true
loss_type: mse
lr: 0.0002
noise_scheduler: flowmatch
optimizer: adamw8bit
reg_weight: 1.0
steps: 3000
target_noise_multiplier: 1.0
train_text_encoder: false
train_unet: true
training_folder: /root/lorahub
trigger_word: in the style of dstyl3xl
type: sd_trainer
job: extension
meta:
description: Trained on a set of MidJourney v6 images made using a fixed blend
of srefs and image prompts, going for a vivid 70's illustration style