Model card auto-generated by SimpleTuner

README.md

@@ -18,7 +18,7 @@ widget:
     negative_prompt: 'blurry, cropped, ugly'
   output:
     url: ./assets/image_0_0.png
-- text: 'The building exhibits a minimalist architectural approach featuring a stark, white cubic form. Sharp, clean lines define its structure, enhancing a geometric precision. The facade is punctuated by rectangular windows and openings, interspersed asymmetrically. The smooth white stucco finish reflects natural light, creating a play of light and shadow that enhances the purity of its form. Surrounded by a similarly white, flat-roofed wall that delineates the boundary without detracting from the building’s monolithic appearance, the surrounding environment features some neighboring residences with tiled roofs and contemporary designs, but the focus remains on the striking simplicity and modern aesthetics of the cube structure against a clear blue sky.'
+- text: 'modern architecture, canopy structures, white material, urban design, outdoor space, trees, landscaping, seating areas, people, daylight, clear sky, recreational area, paving pattern, public area, contemporary design, pergola-like elements, radial pattern, greenery, mixed-use space'
   parameters:
     negative_prompt: 'blurry, cropped, ugly'
   output:
@@ -31,20 +31,19 @@ This is a LyCORIS adapter derived from [black-forest-labs/FLUX.1-dev](https://hu
 
 
 The main validation prompt used during training was:
-
-
-
 ```
-The building exhibits a minimalist architectural approach featuring a stark, white cubic form. Sharp, clean lines define its structure, enhancing a geometric precision. The facade is punctuated by rectangular windows and openings, interspersed asymmetrically. The smooth white stucco finish reflects natural light, creating a play of light and shadow that enhances the purity of its form. Surrounded by a similarly white, flat-roofed wall that delineates the boundary without detracting from the building’s monolithic appearance, the surrounding environment features some neighboring residences with tiled roofs and contemporary designs, but the focus remains on the striking simplicity and modern aesthetics of the cube structure against a clear blue sky.
+modern architecture, canopy structures, white material, urban design, outdoor space, trees, landscaping, seating areas, people, daylight, clear sky, recreational area, paving pattern, public area, contemporary design, pergola-like elements, radial pattern, greenery, mixed-use space
 ```
 
+
 ## Validation settings
 - CFG: `3.0`
 - CFG Rescale: `0.0`
 - Steps: `20`
-- Sampler: `None`
+- Sampler: `FlowMatchEulerDiscreteScheduler`
 - Seed: `42`
 - Resolution: `1024x1024`
+- Skip-layer guidance: 
 
 Note: The validation settings are not necessarily the same as the [training settings](#training-settings).
 
@@ -59,20 +58,24 @@ You may reuse the base model text encoder for inference.
 
 ## Training settings
 
-- Training epochs: 2
-- Training steps: 1000
+- Training epochs: 0
+- Training steps: 500
 - Learning rate: 0.0001
+  - Learning rate schedule: polynomial
+  - Warmup steps: 100
+- Max grad norm: 2.0
 - Effective batch size: 1
   - Micro-batch size: 1
   - Gradient accumulation steps: 1
   - Number of GPUs: 1
-- Prediction type: flow-matching
-- Rescaled betas zero SNR: False
+- 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
-- Precision: Pure BF16
-- Quantised: No
-- Xformers: Not used
-- LyCORIS Config:
+- Trainable parameter precision: Pure BF16
+- Caption dropout probability: 5.0%
+
+
+### LyCORIS Config:
 ```json
 {
     "algo": "lokr",
@@ -99,14 +102,96 @@ You may reuse the base model text encoder for inference.
 
 ## Datasets
 
+### my-dataset-256
+- Repeats: 10
+- Total number of images: 71
+- Total number of aspect buckets: 1
+- Resolution: 0.065536 megapixels
+- Cropped: False
+- Crop style: None
+- Crop aspect: None
+- Used for regularisation data: No
+### my-dataset-crop-256
+- Repeats: 10
+- Total number of images: 71
+- Total number of aspect buckets: 1
+- Resolution: 0.065536 megapixels
+- Cropped: True
+- Crop style: center
+- Crop aspect: square
+- Used for regularisation data: No
+### my-dataset-512
+- Repeats: 10
+- Total number of images: 71
+- Total number of aspect buckets: 1
+- Resolution: 0.262144 megapixels
+- Cropped: False
+- Crop style: None
+- Crop aspect: None
+- Used for regularisation data: No
+### my-dataset-crop-512
+- Repeats: 10
+- Total number of images: 71
+- Total number of aspect buckets: 1
+- Resolution: 0.262144 megapixels
+- Cropped: True
+- Crop style: center
+- Crop aspect: square
+- Used for regularisation data: No
+### my-dataset-768
+- Repeats: 10
+- Total number of images: 71
+- Total number of aspect buckets: 1
+- Resolution: 0.589824 megapixels
+- Cropped: False
+- Crop style: None
+- Crop aspect: None
+- Used for regularisation data: No
+### my-dataset-crop-768
+- Repeats: 10
+- Total number of images: 71
+- Total number of aspect buckets: 1
+- Resolution: 0.589824 megapixels
+- Cropped: True
+- Crop style: center
+- Crop aspect: square
+- Used for regularisation data: No
 ### my-dataset-1024
-- Repeats: 1
-- Total number of images: 209
-- Total number of aspect buckets: 4
+- Repeats: 10
+- Total number of images: 71
+- Total number of aspect buckets: 1
 - Resolution: 1.048576 megapixels
 - Cropped: False
 - Crop style: None
 - Crop aspect: None
+- Used for regularisation data: No
+### my-dataset-crop-1024
+- Repeats: 10
+- Total number of images: 71
+- Total number of aspect buckets: 1
+- Resolution: 1.048576 megapixels
+- Cropped: True
+- Crop style: center
+- Crop aspect: square
+- Used for regularisation data: No
+### my-dataset-1440
+- Repeats: 10
+- Total number of images: 71
+- Total number of aspect buckets: 1
+- Resolution: 2.0736 megapixels
+- Cropped: False
+- Crop style: None
+- Crop aspect: None
+- Used for regularisation data: No
+### my-dataset-crop-1440
+- Repeats: 10
+- Total number of images: 71
+- Total number of aspect buckets: 1
+- Resolution: 2.0736 megapixels
+- Cropped: True
+- Crop style: center
+- Crop aspect: square
+- Used for regularisation data: No
 
 
 ## Inference
@@ -117,19 +202,45 @@ 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_id = 'pytorch_lora_weights.safetensors' # you will have to download this manually
+adapter_repo_id = 'ossaili/simpletuner-lora'
+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_id, pipeline.transformer)
+wrapper, _ = create_lycoris_from_weights(lora_scale, adapter_file_path, pipeline.transformer)
 wrapper.merge_to()
 
-prompt = "The building exhibits a minimalist architectural approach featuring a stark, white cubic form. Sharp, clean lines define its structure, enhancing a geometric precision. The facade is punctuated by rectangular windows and openings, interspersed asymmetrically. The smooth white stucco finish reflects natural light, creating a play of light and shadow that enhances the purity of its form. Surrounded by a similarly white, flat-roofed wall that delineates the boundary without detracting from the building’s monolithic appearance, the surrounding environment features some neighboring residences with tiled roofs and contemporary designs, but the focus remains on the striking simplicity and modern aesthetics of the cube structure against a clear blue sky."
+prompt = "modern architecture, canopy structures, white material, urban design, outdoor space, trees, landscaping, seating areas, people, daylight, clear sky, recreational area, paving pattern, public area, contemporary design, pergola-like elements, radial pattern, greenery, mixed-use space"
+
 
-pipeline.to('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
+## 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(1641421826),
+    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,
@@ -137,3 +248,5 @@ image = pipeline(
 image.save("output.png", format="PNG")
 ```
 
+
+