Merge branch 'ClownsharkBatwing:main' into main

conditioning.py

@@ -446,7 +446,7 @@ def __call__(self, transformer_options, threshold, threshold_inv=False, dtype=to
             if self.mask_type == "gradient":
                 return self.mask.clone().to(sigma.device).to(dtype)
             elif self.mask_type == "differential":
-                return self.mask.clone().to(sigma.device) >= threshold
+                return self.mask.clone().to(sigma.device) > threshold
                 if threshold_inv==False:
                     mask_tmp = self.mask.to(dtype).clone().to(sigma.device) >= threshold
                     return mask_tmp

flux/model.py

@@ -96,16 +96,27 @@ def forward_blocks(self, img: Tensor, img_ids: Tensor, txt: Tensor, txt_ids: Ten
         pe = self.pe_embedder(ids)
 
         weight = transformer_options['reg_cond_weight'] if 'reg_cond_weight' in transformer_options else 0.0
+        threshold_factor_double = transformer_options['reg_cond_diff_threshold_factor_double'] if 'reg_cond_diff_threshold_factor_double' in transformer_options else 1.0
+        threshold_factor_single = transformer_options['reg_cond_diff_threshold_factor_single'] if 'reg_cond_diff_threshold_factor_single' in transformer_options else 0.5
+
+        threshold_factor_double_absolute = transformer_options['reg_cond_diff_threshold_factor_double_absolute'] if 'reg_cond_diff_threshold_factor_double_absolute' in transformer_options else -1
+        threshold_factor_single_absolute = transformer_options['reg_cond_diff_threshold_factor_single_absolute'] if 'reg_cond_diff_threshold_factor_single_absolute' in transformer_options else -1
 
         for i, block in enumerate(self.double_blocks):
             mask = None
             mask_obj = transformer_options.get('patches', {}).get('regional_conditioning_mask', None)
-            threshold = i / 56
+            threshold = i / 57
             if mask_obj is not None and weight >= threshold:
                 #if self.threshold_inv:
                 #    threshold = 1 - threshold
-                mask = mask_obj[0](transformer_options, threshold) #, self.threshold_inv)
+                #threshold = (56-i)/57
+                if threshold_factor_double_absolute < 0:
+                    mask_threshold = threshold * threshold_factor_double
+                else: 
+                    mask_threshold = threshold_factor_double_absolute
+                mask = mask_obj[0](transformer_options, mask_threshold) #, self.threshold_inv)
                 #mask = (mask >= threshold).to(mask.dtype)
+            #print("i = ", i, threshold, weight)
 
             img, txt = block(img=img, txt=txt, vec=vec, pe=pe, timestep=timesteps, transformer_options=transformer_options, mask=mask) #, mask=mask)
 
@@ -120,13 +131,19 @@ def forward_blocks(self, img: Tensor, img_ids: Tensor, txt: Tensor, txt_ids: Ten
         for i, block in enumerate(self.single_blocks):
             mask = None
             mask_obj = transformer_options.get('patches', {}).get('regional_conditioning_mask', None)
-            threshold = (1+18)/56
+            threshold = (i+18)/57
             if mask_obj is not None and weight >= threshold:
                 #if self.threshold_inv:
                 #    threshold = 1 - threshold
-                mask = mask_obj[0](transformer_options, threshold) #, self.threshold_inv)
+                #threshold = (56-(i+18))/57
+                if threshold_factor_single_absolute < 0:
+                    mask_threshold = threshold * threshold_factor_single
+                else: 
+                    mask_threshold = threshold_factor_single_absolute
+                mask = mask_obj[0](transformer_options, mask_threshold) #, self.threshold_inv)
                 #threshold = (i+18)/56
                 #mask = (mask >= threshold).to(mask.dtype)
+            #print("i2 = ", i, threshold, weight)
 
             img = block(img, vec=vec, pe=pe, timestep=timesteps, transformer_options=transformer_options, mask=mask)
 

rk_sampler.py

@@ -145,6 +145,10 @@ def post_cfg_function(args):
         model_options = extra_args.get("model_options", {}).copy()
         extra_args["model_options"] = comfy.model_patcher.set_model_options_post_cfg_function(model_options, post_cfg_function, disable_cfg1_optimization=True)  
 
+    extra_args['model_options']['transformer_options']['reg_cond_diff_threshold_factor_double'] = float(get_extra_options_kv("reg_cond_diff_threshold_factor_double", "0.5", extra_options))
+    extra_args['model_options']['transformer_options']['reg_cond_diff_threshold_factor_single'] = float(get_extra_options_kv("reg_cond_diff_threshold_factor_single", "0.5", extra_options))
+    extra_args['model_options']['transformer_options']['reg_cond_diff_threshold_factor_double_absolute'] = float(get_extra_options_kv("reg_cond_diff_threshold_factor_double_absolute", "-1", extra_options))
+    extra_args['model_options']['transformer_options']['reg_cond_diff_threshold_factor_single_absolute'] = float(get_extra_options_kv("reg_cond_diff_threshold_factor_single_absolute", "-1", extra_options))
     if extra_options_flag("cfg_cw", extra_options):
         cfg_cw = float(get_extra_options_kv("cfg_cw", "1.0", extra_options))
     extra_args = rk.init_cfg_channelwise(x, cfg_cw, **extra_args)
@@ -162,6 +166,7 @@ def post_cfg_function(args):
         else:
             #model.inner_model.model_options['transformer_options']['patches']['regional_conditioning_weight'] = 0.0
             extra_args['model_options']['transformer_options']['regional_conditioning_weight'] = 0.0
+
         eta = eta_var = etas[step] if etas is not None else eta
         s_noise = s_noises[step] if s_noises is not None else s_noise
 

samplers.py

@@ -103,6 +103,7 @@ def main(self, model, cfg, sampler_mode, scheduler, steps, denoise=1.0, denoise_
                     model.set_model_patch(regional_mask,         'regional_conditioning_mask')
 
             if "extra_options" in sampler.extra_options:
+                extra_options += " "
                 extra_options += sampler.extra_options['extra_options']
                 sampler.extra_options['extra_options'] = extra_options