metadata
license: mit
base_model: black-forest-labs/FLUX.1-dev
tags:
- flux
- flux-diffusers
- text-to-image
- diffusers
- simpletuner
- lora
- lycoris
- template:sd-lora
widget:
- text: >-
A towering colossus of rusted metal and stone looms over a battlefield,
its hollow chest holding a flickering, eldritch flame. Around its feet,
fallen soldiers rise once more, bound by its curse.
output:
url: images/gothic_001.jpg
- text: >-
A dragon with scales like liquid mercury coils around a shattered castle
spire, its many eyes reflecting countless worlds.
output:
url: images/gothic_002.jpg
- text: >-
A skeletal monarch clad in decaying finery sits on a throne made of
cracked bones, his hollow eyes glowing faintly. Behind him, tattered
banners whisper in an invisible wind.
output:
url: images/gothic_003.jpg
- text: >-
An assassin cloaked in shadows, his face obscured by a hood, in the dimly
lit alleyways of an ancient city. his dagger gleams with a dark liquid,
and his eyes scan the darkness for his prey.
output:
url: images/gothic_004.jpg
Training settings
- Training steps: 1900
- Learning rate: 2e-4
- Effective batch size: 6
- Micro-batch size: 6
- Gradient accumulation steps: 1
- Number of GPUs: 1
- Prediction type: flow-matching
- Rescaled betas zero SNR: False
- Optimizer: adamw_bf16
- Precision: Pure BF16
- Gradient Precision: FP32
- Lora type: lycoris