flux hires and face-hires

CHANGELOG.md

@@ -7,7 +7,7 @@
 Major refactor of [FLUX.1](https://blackforestlabs.ai/announcing-black-forest-labs/) support:
 - Full **ControlNet** support, better **LoRA** support, full **prompt attention** support,
 - Faster execution, more flexible loading, additional quantization options, and more...
-- Added **image-to-image**, **inpaint** and **outpaint** modes  
+- Added **image-to-image**, **inpaint**, **outpaint**, **hires** modes  
 - Since both *Optimum-Quanto* and *BitsAndBytes* libraries are limited in their platform support matrix,  
   try enabling **NNCF** for quantization/compression on-the-fly!
 
@@ -50,7 +50,8 @@ Plus tons of minor items and fixes - see [changelog](https://github.com/vladmand
   this brings supported quants to: *nf4/fp8/fp4/qint8/qint4*
 - vae support *fp16*  
 - **lora** support additional training tools  
-- support fuse-qkv projections  
+- **face-hires** support
+- support **fuse-qkv** projections  
   can speed up generate  
   enable via *settings -> compute -> fused projections*  
 

modules/modeldata.py

@@ -103,9 +103,9 @@ class Shared(sys.modules[__name__].__class__):
                 model_type = 'sc'
             elif "AuraFlow" in self.sd_model.__class__.__name__:
                 model_type = 'auraflow'
-            elif "FluxPipeline" in self.sd_model.__class__.__name__ or "FluxControlNetPipeline" in self.sd_model.__class__.__name__:
+            elif "Flux" in self.sd_model.__class__.__name__:
                 model_type = 'f1'
-            elif "CogVideoXPipeline" in self.sd_model.__class__.__name__ or "CogVideoXVideoToVideoPipeline":
+            elif "CogVideo" in self.sd_model.__class__.__name__:
                 model_type = 'cogvideox'
             else:
                 model_type = self.sd_model.__class__.__name__

modules/sd_samplers.py

@@ -71,7 +71,7 @@ def create_sampler(name, model):
         sampler = config.constructor(model)
         if shared.sd_model_type == 'f1':
             if 'base_image_seq_len' not in sampler.sampler.config or 'max_image_seq_len' not in sampler.sampler.config or 'base_shift' not in sampler.sampler.config or 'max_shift' not in sampler.sampler.config:
-                shared.log.warning(f'FLUX: sampler="{name}" non compatible')
+                shared.log.warning(f'FLUX: sampler="{name}" unsupported')
                 # sampler.sampler.register_to_config(base_image_seq_len=256, max_image_seq_len=4096, base_shift=0.5, max_shift=1.15)
                 return None
         if not hasattr(model, 'scheduler_config'):