--- license: other base_model: "black-forest-labs/FLUX.1-dev" tags: - flux - flux-diffusers - text-to-image - diffusers - simpletuner - safe-for-work - lora - template:sd-lora - lycoris inference: true widget: - text: 'unconditional (blank prompt)' parameters: negative_prompt: 'blurry, cropped, ugly' output: url: ./assets/image_0_0.png - text: 'In the style of Frank C. Pape fairy tale illustrations, A warrior princess in flowing silver armor rides a white horse through falling snow, her long cape billowing behind her. She holds a glowing crystal staff while three ravens circle overhead near a stone archway.' parameters: negative_prompt: 'blurry, cropped, ugly' output: url: ./assets/image_1_0.png - text: 'In the style of Frank C. Pape fairy tale illustrations, A bearded wizard in a star-patterned robe stands atop a rocky cliff, raising his hands toward storm clouds while ships with golden sails battle waves below. Sea creatures with gleaming scales leap from the turbulent waters.' parameters: negative_prompt: 'blurry, cropped, ugly' output: url: ./assets/image_2_0.png - text: 'In the style of Frank C. Pape fairy tale illustrations, A woman in an emerald dress with intricate gold embroidery sits beneath a flowering tree, offering a silver goblet to a deer. In the background, a castle with twisted spires rises against a sunset sky.' parameters: negative_prompt: 'blurry, cropped, ugly' output: url: ./assets/image_3_0.png - text: 'In the style of Frank C. Pape fairy tale illustrations, A giant golden hamster wearing burnished steel armor and a crimson velvet cape sits upon an ornate throne carved from ancient oak and golden wheat. Mice in blue and silver livery bow before him, presenting jeweled acorns on silk cushions while court musicians play tiny silver trumpets.' parameters: negative_prompt: 'blurry, cropped, ugly' output: url: ./assets/image_4_0.png - text: 'In the style of Frank C. Pape fairy tale illustrations, A mysterious merchant in an emerald robe and golden mask holds up a glowing Coca-Cola bottle beneath a canopy of twisted oak branches. Forest creatures in medieval dress gather around its ruby light, while silver-winged fairies dance through moonbeams that filter through the leaves.' parameters: negative_prompt: 'blurry, cropped, ugly' output: url: ./assets/image_5_0.png - text: 'In the style of Frank C. Pape fairy tale illustrations, A Range Rover with brass-and-silver clockwork wheels and gleaming armor plates crosses an ancient stone bridge. Four mechanical horses with steam-breathing nostrils and copper manes pull it through swirling silver mist, while a wizard in a pinstripe suit raises a crystal staff from the driver''s seat.' parameters: negative_prompt: 'blurry, cropped, ugly' output: url: ./assets/image_6_0.png - text: 'In the style of Frank C. Pape fairy tale illustrations, A sorcerer in purple silk robes trimmed with gold stands atop a winding stone staircase, conducting floating books with a feather quill that trails sparks. Beneath gothic arches, apprentices in pointed hats ride enchanted carpets between towering bookshelves of ancient tomes.' parameters: negative_prompt: 'blurry, cropped, ugly' output: url: ./assets/image_7_0.png - text: 'In the style of Frank C. Pape fairy tale illustrations, A grand feast hall with tapestry-hung walls where animal nobles in velvet and silk dine at a table of polished oak. At its center, a towering crystal fountain flows with sparkling Coca-Cola, while rabbit jesters in bells and motley juggle glowing bottles beneath chandeliers.' parameters: negative_prompt: 'blurry, cropped, ugly' output: url: ./assets/image_8_0.png --- # FrankPape-RussianStoryBook-Flux-LoKr-4e-4 This is a LyCORIS adapter derived from [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev). No validation prompt was used during training. None ## Validation settings - CFG: `3.0` - CFG Rescale: `0.0` - Steps: `20` - Sampler: `FlowMatchEulerDiscreteScheduler` - Seed: `42` - Resolution: `1024x1024` - Skip-layer guidance: Note: The validation settings are not necessarily the same as the [training settings](#training-settings). You can find some example images in the following gallery: The text encoder **was not** trained. You may reuse the base model text encoder for inference. ## Training settings - Training epochs: 27 - Training steps: 6000 - Learning rate: 1e-05 - Learning rate schedule: constant - Warmup steps: 200 - Max grad norm: 2.0 - Effective batch size: 3 - Micro-batch size: 3 - Gradient accumulation steps: 1 - Number of GPUs: 1 - Gradient checkpointing: True - Prediction type: flow-matching (extra parameters=['shift=3', 'flux_guidance_mode=constant', 'flux_guidance_value=1.0', 'flow_matching_loss=compatible']) - Optimizer: adamw_bf16 - Trainable parameter precision: Pure BF16 - Caption dropout probability: 10.0% ### LyCORIS Config: ```json { "algo": "lokr", "multiplier": 1.0, "linear_dim": 10000, "linear_alpha": 1, "factor": 16, "apply_preset": { "target_module": [ "Attention", "FeedForward" ], "module_algo_map": { "Attention": { "factor": 16 }, "FeedForward": { "factor": 8 } } } } ``` ## Datasets ### fws-512 - Repeats: 10 - Total number of images: 16 - Total number of aspect buckets: 2 - Resolution: 0.262144 megapixels - Cropped: False - Crop style: None - Crop aspect: None - Used for regularisation data: No ### fws-1024 - Repeats: 6 - Total number of images: 16 - Total number of aspect buckets: 2 - Resolution: 1.048576 megapixels - Cropped: False - Crop style: None - Crop aspect: None - Used for regularisation data: No ### fws-512-crop - Repeats: 10 - Total number of images: 16 - Total number of aspect buckets: 1 - Resolution: 0.262144 megapixels - Cropped: True - Crop style: random - Crop aspect: square - Used for regularisation data: No ### fws-1024-crop - Repeats: 6 - Total number of images: 16 - Total number of aspect buckets: 1 - Resolution: 1.048576 megapixels - Cropped: True - Crop style: random - Crop aspect: square - Used for regularisation data: No ## Inference ```python 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 = 'black-forest-labs/FLUX.1-dev' adapter_repo_id = 'davidrd123/FrankPape-RussianStoryBook-Flux-LoKr-4e-4' 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 = "An astronaut is riding a horse through the jungles of Thailand." ## Optional: quantise the model to save on vram. ## Note: The model was quantised during training, and so it is recommended to do the same 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 image = pipeline( prompt=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(42), width=1024, height=1024, guidance_scale=3.0, ).images[0] image.save("output.png", format="PNG") ``` ## Exponential Moving Average (EMA) SimpleTuner generates a safetensors variant of the EMA weights and a pt file. The safetensors file is intended to be used for inference, and the pt file is for continuing finetuning. The EMA model may provide a more well-rounded result, but typically will feel undertrained compared to the full model as it is a running decayed average of the model weights.