sd35m-cheechandchong-masked-reg-autoShift

This is a LyCORIS adapter derived from stabilityai/stable-diffusion-3.5-medium.

The main validation prompt used during training was:

A photo-realistic image of a cat

Validation settings

• CFG: 3.0
• CFG Rescale: 0.0
• Steps: 20
• Sampler: None
• Seed: 42
• Resolution: 1024x1024

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

a picture of tommy chong

Negative Prompt

blurry, cropped, ugly

Prompt

young tommy chong

Negative Prompt

blurry, cropped, ugly

Prompt

a stoic photograph of tommy chong. he looks off into the distance, standing up against the railing of a ship. the sky is cloudy.

Negative Prompt

blurry, cropped, ugly

Prompt

an elderly tommy chong as a contestant on Wheel of Fortune

Negative Prompt

blurry, cropped, ugly

Prompt

tommy chong as a superhero in the style of studio ghibli. he wears a metal armor suit with glowing lights and power indicators.

Negative Prompt

blurry, cropped, ugly

Prompt

tommy chong in a casket, dead. he is dead and it is a funeral. the text overhead says 'HE HAS NOT RISEN'.

Negative Prompt

blurry, cropped, ugly

Prompt

a picture of cheech marin

Negative Prompt

blurry, cropped, ugly

Prompt

young cheech marin

Negative Prompt

blurry, cropped, ugly

Prompt

a stoic photograph of cheech marin. he looks off into the distance, standing up against the railing of a ship. the sky is cloudy.

Negative Prompt

blurry, cropped, ugly

Prompt

an elderly cheech marin as a contestant on Wheel of Fortune

Negative Prompt

blurry, cropped, ugly

Prompt

cheech marin as a superhero in the style of studio ghibli. he wears a metal armor suit with glowing lights and power indicators.

Negative Prompt

blurry, cropped, ugly

Prompt

cheech marin in a casket, dead. he is dead and it is a funeral. the text overhead says 'HE HAS NOT RISEN'.

Negative Prompt

blurry, cropped, ugly

Prompt

cheech marin sitting to the left of tommy chong on the set of a television interview

Negative Prompt

blurry, cropped, ugly

Prompt

cheech marin sitting to the right of tommy chong on the set of a television interview

Negative Prompt

blurry, cropped, ugly

Prompt

cheech and chong sitting together on the stoop of a new york apartment building, 1972

Negative Prompt

blurry, cropped, ugly

Prompt

the iconic duo cheech and chong on stage performing stand-up comedy together in 2008

Negative Prompt

blurry, cropped, ugly

Prompt

A photo-realistic image of a cat

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: 15
• Training steps: 5000
• Learning rate: 0.0001
• Max grad norm: 0.01
• Effective batch size: 12
  • Micro-batch size: 4
  • Gradient accumulation steps: 1
  • Number of GPUs: 3
• Prediction type: flow-matching
• Rescaled betas zero SNR: False
• Optimizer: bnb-adamw8bit
• Precision: Pure BF16
• Quantised: No
• Xformers: Not used
• LyCORIS Config:

{
    "bypass_mode": true,
    "algo": "lokr",
    "multiplier": 1.0,
    "full_matrix": true,
    "linear_dim": 10000,
    "linear_alpha": 1,
    "factor": 12,
    "apply_preset": {
        "target_module": [
            "JointTransformerBlock"
        ],
        "module_algo_map": {
            "FeedForward": {
                "factor": 6
            },
            "JointTransformerBlock": {
                "factor": 12
            }
        }
    }
}

Datasets

reg-512

• Repeats: 0
• Total number of images: ~288
• Total number of aspect buckets: 8
• Resolution: 0.262144 megapixels
• Cropped: False
• Crop style: None
• Crop aspect: None
• Used for regularisation data: Yes

reg-1024

• Repeats: 0
• Total number of images: ~291
• Total number of aspect buckets: 3
• Resolution: 1.048576 megapixels
• Cropped: False
• Crop style: None
• Crop aspect: None
• Used for regularisation data: Yes

cheechandchong-uncropped-512

• Repeats: 10
• Total number of images: ~24
• Total number of aspect buckets: 5
• Resolution: 0.262144 megapixels
• Cropped: False
• Crop style: None
• Crop aspect: None
• Used for regularisation data: No

cheechandchong-cropped-512

• Repeats: 10
• Total number of images: ~24
• Total number of aspect buckets: 5
• Resolution: 0.262144 megapixels
• Cropped: False
• Crop style: None
• Crop aspect: None
• Used for regularisation data: No

cheechandchong-uncropped-1024

• Repeats: 10
• Total number of images: ~24
• Total number of aspect buckets: 7
• Resolution: 1.048576 megapixels
• Cropped: False
• Crop style: None
• Crop aspect: None
• Used for regularisation data: No

cheechandchong-cropped-1024

• Repeats: 10
• Total number of images: ~24
• Total number of aspect buckets: 7
• Resolution: 1.048576 megapixels
• Cropped: False
• Crop style: None
• Crop aspect: None
• Used for regularisation data: No

Inference

import torch
from diffusers import DiffusionPipeline
from lycoris import create_lycoris_from_weights

model_id = 'stabilityai/stable-diffusion-3.5-medium'
adapter_id = 'pytorch_lora_weights.safetensors' # you will have to download this manually
lora_scale = 1.0
wrapper, _ = create_lycoris_from_weights(lora_scale, adapter_id, pipeline.transformer)
wrapper.merge_to()

prompt = "A photo-realistic image of a cat"
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=negative_prompt,
    num_inference_steps=20,
    generator=torch.Generator(device='cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu').manual_seed(1641421826),
    width=1024,
    height=1024,
    guidance_scale=3.0,
).images[0]
image.save("output.png", format="PNG")