smoldit-base-test

This is a full rank finetune derived from ptx0/terminus-xl-velocity-v2.

The main validation prompt used during training was:

two young girls in a classroom setting appearing surprised or concerned

Validation settings

• CFG: 4.0
• CFG Rescale: 0.7
• Steps: 30
• Sampler: ddpm
• Seed: 420420420
• Resolution: 256

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

You can find some example images in the following gallery:

Prompt

unconditional (blank prompt)

Negative Prompt

blurry, cropped, ugly

Prompt

two young girls in a classroom setting appearing surprised or concerned

Negative Prompt

blurry, cropped, ugly

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

Training settings

• Training epochs: 360
• Training steps: 167000
• Learning rate: 1e-05
• Effective batch size: 16
  • Micro-batch size: 16
  • Gradient accumulation steps: 1
  • Number of GPUs: 1
• Prediction type: v_prediction
• Rescaled betas zero SNR: True
• Optimizer: AdamW, stochastic bf16
• Precision: Pure BF16
• Xformers: Enabled

Datasets

cinemamix-1mp

• Repeats: 0
• Total number of images: 7408
• Total number of aspect buckets: 1
• Resolution: 256 px
• Cropped: True
• Crop style: center
• Crop aspect: square

Inference

import torch
from diffusers import DiffusionPipeline

model_id = 'smoldit-base-test'
pipeline = DiffusionPipeline.from_pretrained(model_id)

prompt = "two young girls in a classroom setting appearing surprised or concerned"
negative_prompt = "blurry, cropped, ugly"

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=4.0,
    guidance_rescale=0.7,
).images[0]
image.save("output.png", format="PNG")