simpletuner

This is a LyCORIS adapter derived from Lightricks/LTX-Video.

The main validation prompt used during training was:

A photo-realistic image of a cat sitting in a field of lavender flowers. The cat is looking at the viewer.

Validation settings

CFG: 4.2
CFG Rescale: 0.0
Steps: 5
Sampler: FlowMatchEulerDiscreteScheduler
Seed: 42
Resolution: 384x256

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 photo-realistic image of a cat sitting in a field of lavender flowers. The cat is looking at the viewer.

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: 0
Training steps: 60
Learning rate: 8e-05
- Learning rate schedule: constant
- Warmup steps: 0
Max grad value: 2.0
Effective batch size: 4
- Micro-batch size: 4
- Gradient accumulation steps: 1
- Number of GPUs: 1
Gradient checkpointing: True
Prediction type: flow-matching (extra parameters=['training_scheduler_timestep_spacing=trailing', 'inference_scheduler_timestep_spacing=trailing'])
Optimizer: optimi-lion
Trainable parameter precision: Pure BF16
Base model precision: no_change
Caption dropout probability: 10.0%

LyCORIS Config:

{
    "algo": "lokr",
    "multiplier": 1.0,
    "linear_dim": 10000,
    "bypass_mode": true,
    "linear_alpha": 1,
    "factor": 16,
    "apply_preset": {
        "target_module": [
            "Attention",
            "FeedForward"
        ],
        "module_algo_map": {
            "Attention": {
                "factor": 16
            },
            "FeedForward": {
                "factor": 8
            }
        }
    }
}

Datasets

image-dataset-384

Repeats: 4
Total number of images: 11
Total number of aspect buckets: 2
Resolution: 0.147456 megapixels
Cropped: False
Crop style: None
Crop aspect: None
Used for regularisation data: No

video-dataset-384

Repeats: 4
Total number of images: 7
Total number of aspect buckets: 1
Resolution: 0.147456 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


def download_adapter(repo_id: str):
    import os
    from huggingface_hub import hf_hub_download
    adapter_filename = "pytorch_lora_weights.safetensors"
    cache_dir = os.environ.get('HF_PATH', os.path.expanduser('~/.cache/huggingface/hub/models'))
    cleaned_adapter_path = repo_id.replace("/", "_").replace("\\", "_").replace(":", "_")
    path_to_adapter = os.path.join(cache_dir, cleaned_adapter_path)
    path_to_adapter_file = os.path.join(path_to_adapter, adapter_filename)
    os.makedirs(path_to_adapter, exist_ok=True)
    hf_hub_download(
        repo_id=repo_id, filename=adapter_filename, local_dir=path_to_adapter
    )

    return path_to_adapter_file
    
model_id = 'Lightricks/LTX-Video'
adapter_repo_id = 'bghira/simpletuner'
adapter_filename = 'pytorch_lora_weights.safetensors'
adapter_file_path = download_adapter(repo_id=adapter_repo_id)
pipeline = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.bfloat16) # loading directly in bf16
lora_scale = 1.0
wrapper, _ = create_lycoris_from_weights(lora_scale, adapter_file_path, pipeline.transformer)
wrapper.merge_to()

prompt = "A photo-realistic image of a cat sitting in a field of lavender flowers. The cat is looking at the viewer."
negative_prompt = 'blurry, cropped, ugly'

## Optional: quantise the model to save on vram.
## Note: The model was not quantised during training, so it is not necessary to quantise it during inference time.
#from optimum.quanto import quantize, freeze, qint8
#quantize(pipeline.transformer, weights=qint8)
#freeze(pipeline.transformer)
    
pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu') # the pipeline is already in its target precision level
model_output = pipeline(
    prompt=prompt,
    negative_prompt=negative_prompt,
    num_inference_steps=5,
    generator=torch.Generator(device='cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu').manual_seed(42),
    width=384,
    height=256,
    guidance_scale=4.2,
).frames[0]

from diffusers.utils.export_utils import export_to_gif
export_to_gif(model_output, "output.gif", fps=25)

bghira
/

simpletuner