add flux.1-kontext-dev

Signed-off-by: Vladimir Mandic <mandic00@live.com>

CHANGELOG.md

@@ -3,11 +3,18 @@
 ## Update for 2025-06-26
 
 - **Models**
+  - [Models Wiki page](https://vladmandic.github.io/sdnext-docs/Models/) is updated will all new models  
   - [nVidia Cosmos-Predict2 T2I](https://research.nvidia.com/labs/dir/cosmos-predict2/) *2B and 14B*  
-    - new foundational model from Nvidia in two variants: small 2B and large 14B
+    - Cosmos-Predict2 T2I is a new foundational model from Nvidia in two variants: small 2B and large 14B
     - available via *networks -> models -> reference*  
+    - *note*: 14B variant is a very large model at 36GB
     - *note*: this is a gated model, you need to [accept terms](https://huggingface.co/nvidia/Cosmos-Predict2-2B-Text2Image) and set your [huggingface token](https://vladmandic.github.io/sdnext-docs/Gated/)  
-  - [Chroma](https://huggingface.co/lodestones/Chroma)  
+  - [Black Forest Labs FLUX.1 Kontext I2I](https://bfl.ai/announcements/flux-1-kontext-dev) *Dev* variant  
+    - FLUX.1-Kontext is a 12B model billion parameter capable of editing images based on text instructions  
+    - requirements are similar to regular FLUX.1 although 2x slower  
+    - available via *networks -> models -> reference*  
+    - *note*: this is a gated model, you need to [accept terms](https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev) and set your [huggingface token](https://vladmandic.github.io/sdnext-docs/Gated/)  
+  - [lodestones Chroma](https://huggingface.co/lodestones/Chroma)  
     - Chroma is a 8.9B parameter model based on *FLUX.1-schnell* and fully Apache 2.0 licensed  
     - available via *networks -> models -> reference*  
     - *note*: model is still in training so future updates will trigger re-download  

html/reference.json

@@ -158,24 +158,10 @@
     "skip": true,
     "extras": "sampler: Default, cfg_scale: 3.5"
   },
-  "Black Forest Labs FLUX.1 Dev qint8": {
-    "path": "Disty0/FLUX.1-dev-qint8",
-    "preview": "black-forest-labs--FLUX.1-dev.jpg",
-    "desc": "FLUX.1 models are based on a hybrid architecture of multimodal and parallel diffusion transformer blocks, scaled to 12B parameters and builing on flow matching",
-    "skip": true,
-    "extras": "sampler: Default, cfg_scale: 3.5"
-  },
-  "Black Forest Labs FLUX.1 Dev qint4": {
-    "path": "Disty0/FLUX.1-dev-qint4",
-    "preview": "black-forest-labs--FLUX.1-dev.jpg",
-    "desc": "FLUX.1 models are based on a hybrid architecture of multimodal and parallel diffusion transformer blocks, scaled to 12B parameters and builing on flow matching",
-    "skip": true,
-    "extras": "sampler: Default, cfg_scale: 3.5"
-  },
-  "Black Forest Labs FLUX.1 Dev nf4": {
-    "path": "sayakpaul/flux.1-dev-nf4",
-    "preview": "black-forest-labs--FLUX.1-dev.jpg",
-    "desc": "FLUX.1 models are based on a hybrid architecture of multimodal and parallel diffusion transformer blocks, scaled to 12B parameters and builing on flow matching",
+  "Black Forest Labs FLUX.1 Kontext Dev": {
+    "path": "black-forest-labs/FLUX.1-Kontext-dev",
+    "preview": "black-forest-labs--FLUX.1-Kontext-dev.jpg",
+    "desc": "FLUX.1 Kontext [dev] is a 12 billion parameter rectified flow transformer capable of editing images based on text instructions.",
     "skip": true,
     "extras": "sampler: Default, cfg_scale: 3.5"
   },

modules/model_flux.py

@@ -317,6 +317,13 @@ def load_flux(checkpoint_info, diffusers_load_config): # triggered by opts.sd_ch
         cls = diffusers.FluxControlPipeline
     elif 'Depth' in repo_id:
         cls = diffusers.FluxControlPipeline
+    elif 'Kontext' in repo_id:
+        cls = diffusers.FluxKontextPipeline
+        from diffusers import pipelines
+        pipelines.auto_pipeline.AUTO_TEXT2IMAGE_PIPELINES_MAPPING["flux1kontext"] = diffusers.FluxKontextPipeline
+        pipelines.auto_pipeline.AUTO_IMAGE2IMAGE_PIPELINES_MAPPING["flux1kontext"] = diffusers.FluxKontextPipeline
+        pipelines.auto_pipeline.AUTO_INPAINT_PIPELINES_MAPPING["flux1kontext"] = diffusers.FluxKontextPipeline
+
     else:
         cls = diffusers.FluxPipeline
     shared.log.debug(f'Load model: type=FLUX cls={cls.__name__} preloaded={list(kwargs)} revision={diffusers_load_config.get("revision", None)}')

modules/processing_args.py

@@ -48,17 +48,21 @@ def task_specific_kwargs(p, model):
             'image': p.init_images,
             'strength': p.denoising_strength,
         }
-        if model.__class__.__name__ == 'FluxImg2ImgPipeline': # needs explicit width/height
+        if model.__class__.__name__ == 'FluxImg2ImgPipeline' or model.__class__.__name__ == 'FluxKontextPipeline': # needs explicit width/height
             if torch.is_tensor(p.init_images[0]):
-                p.width = p.init_images[0].shape[-1] * 16
-                p.height = p.init_images[0].shape[-2] * 16
+                p.width, p.height = p.init_images[0].shape[-1] * 16, p.init_images[0].shape[-2] * 16
             else:
-                p.width = 8 * math.ceil(p.init_images[0].width / 8)
-                p.height = 8 * math.ceil(p.init_images[0].height / 8)
+                p.width, p.height = 8 * math.ceil(p.init_images[0].width / 8), 8 * math.ceil(p.init_images[0].height / 8)
+            if model.__class__.__name__ == 'FluxKontextPipeline':
+                aspect_ratio = p.width / p.height
+                vae_scale_factor = 16
+                max_area = max(p.width, p.height)**2
+                p.width, p.height = round((max_area * aspect_ratio) ** 0.5), round((max_area / aspect_ratio) ** 0.5)
+                p.width, p.height = p.width // vae_scale_factor * vae_scale_factor, p.height // vae_scale_factor * vae_scale_factor
+                task_args['max_area'] = max_area
             task_args['width'], task_args['height'] = p.width, p.height
         if model.__class__.__name__ == 'OmniGenPipeline':
-            p.width = 16 * math.ceil(p.init_images[0].width / 16)
-            p.height = 16 * math.ceil(p.init_images[0].height / 16)
+            p.width, p.height = 16 * math.ceil(p.init_images[0].width / 16), 16 * math.ceil(p.init_images[0].height / 16)
             task_args = {
                 'width': p.width,
                 'height': p.height,

modules/processing_helpers.py

@@ -466,9 +466,7 @@ def calculate_base_steps(p, use_denoise_start, use_refiner_start):
         cls = shared.sd_model.__class__.__name__
         if cls in sd_models.i2i_pipes:
             steps = p.steps
-        elif 'Flex' in cls:
-            steps = p.steps
-        elif 'HiDreamImageEditingPipeline' in cls:
+        elif 'Flex' in cls or 'HiDreamImageEditingPipeline' in cls or 'Kontext' in cls:
             steps = p.steps
         elif use_denoise_start and (shared.sd_model_type == 'sdxl'):
             steps = p.steps // (1 - p.refiner_start)

modules/ui_models.py

@@ -37,7 +37,14 @@ def create_ui():
                     model = modelstats.analyze()
                     desc = f"Model: {model.name}<br>Type: {model.type}<br>Class: {model.cls}<br>Size: {model.size} bytes<br>Modified: {model.mtime}<br>"
                     meta = model.meta
-                    components = [(m.name, m.cls, m.device, m.dtype, m.params, m.modules, str(m.config)) for m in model.modules]
+                    components = []
+                    for m in model.modules:
+                        try:
+                            component = (m.name, m.cls, str(m.device), str(m.dtype), m.params, m.modules, str(m.config))
+                            components.append(component)
+                        except Exception:
+                            component = (m.name, m.cls, str(m.device), str(m.dtype), m.params, m.modules, '')
+                            components.append(component)
                     return [desc, components, meta]
 
                 with gr.Row():