terminus-xl-refiner

This is a full rank finetune derived from segmind/SSD-1B.

The main validation prompt used during training was:

a cute anime character named toast

Validation settings

• CFG: 7.5
• CFG Rescale: 0.7
• Steps: 30
• Sampler: ddpm
• Seed: 420420420
• Resolution: 1024

Note: The validation settings are not necessarily the same as the training settings.

The text encoder was not trained. You may reuse the base model text encoder for inference.

Training settings

• Training epochs: 0
• Training steps: 12800
• Learning rate: 2e-06
• Effective batch size: 16
  • Micro-batch size: 4
  • Gradient accumulation steps: 4
  • Number of GPUs: 1
• Prediction type: v_prediction
• Rescaled betas zero SNR: True
• Optimizer: AdamW, stochastic bf16
• Precision: Pure BF16
• Xformers: Enabled

Datasets

pixel-art

• Repeats: 0
• Total number of images: 1040
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

signs

• Repeats: 0
• Total number of images: 368
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

experimental

• Repeats: 0
• Total number of images: 3024
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

ethnic

• Repeats: 0
• Total number of images: 3072
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

sports

• Repeats: 0
• Total number of images: 784
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

architecture

• Repeats: 0
• Total number of images: 4336
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

shutterstock

• Repeats: 0
• Total number of images: 21072
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

cinemamix-1mp

• Repeats: 0
• Total number of images: 9008
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

nsfw-1024

• Repeats: 0
• Total number of images: 10800
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

anatomy

• Repeats: 5
• Total number of images: 16417
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

yoga

• Repeats: 0
• Total number of images: 3600
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

photo-aesthetics

• Repeats: 0
• Total number of images: 33136
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

text-1mp

• Repeats: 5
• Total number of images: 13170
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

photo-concept-bucket

• Repeats: 0
• Total number of images: 567554
• Total number of aspect buckets: 3
• Resolution: 1.0 megapixels
• Cropped: True
• Crop style: random
• Crop aspect: random

Inference

import torch
from diffusers import DiffusionPipeline



model_id = "terminus-xl-refiner"
prompt = "a cute anime character named toast"
negative_prompt = "malformed, disgusting, overexposed, washed-out"

pipeline = DiffusionPipeline.from_pretrained(model_id)
pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
image = pipeline(
    prompt=prompt,
    negative_prompt='blurry, cropped, ugly',
    num_inference_steps=30,
    generator=torch.Generator(device='cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu').manual_seed(1641421826),
    width=1152,
    height=768,
    guidance_scale=7.5,
    guidance_rescale=0.7,
).images[0]
image.save("output.png", format="PNG")