keras diffusers stable-diffusion text-to-image keras-dreambooth consentful

Model description

Model description

This model has been fine-tuned to learn the concept of Hokusai artist by ChienVM on this dataset

Intended uses & limitations

Concept token "hks## style". Example: "a photo of painting image in hks## style"

Some examples prompt

<figure> <img src="https://huggingface.co/keras-dreambooth/dreambooth_diffusion_hokusai/resolve/main/image_01.png" alt="Sample 1"> <figcaption>Prompt: a painting image of a fishing village under a cherry blossom forest at sunset in hks## style, ultra realistic, 4k, 8k. Negative prompt: bad anatomy, soft blurry. Steps: 100, CFG scale: 15, Size: 512x12 </figcaption> </figure>

<figure> <img src="https://huggingface.co/keras-dreambooth/dreambooth_diffusion_hokusai/resolve/main/image_02.png" alt="Sample 2"> <figcaption>Prompt: a beautiful and highly detailed oil painting of a lost valley in the mountains in hks## style, running river, intricate details, 8k, sharp focus, hyper realism. Negative prompt: (bad anatomy), (blurry), grain. Steps: 100, CFG scale: 15, Size: 512x12 </figcaption> </figure>

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

Hyperparameters Value
inner_optimizer.class_name Custom>RMSprop
inner_optimizer.config.name RMSprop
inner_optimizer.config.weight_decay None
inner_optimizer.config.clipnorm None
inner_optimizer.config.global_clipnorm None
inner_optimizer.config.clipvalue None
inner_optimizer.config.use_ema False
inner_optimizer.config.ema_momentum 0.99
inner_optimizer.config.ema_overwrite_frequency 100
inner_optimizer.config.jit_compile True
inner_optimizer.config.is_legacy_optimizer False
inner_optimizer.config.learning_rate 0.0010000000474974513
inner_optimizer.config.rho 0.9
inner_optimizer.config.momentum 0.0
inner_optimizer.config.epsilon 1e-07
inner_optimizer.config.centered False
dynamic True
initial_scale 32768.0
dynamic_growth_steps 2000
training_precision mixed_float16