Merge pull request #4021 from AUTOMATIC1111/add-kdiff-cfgdenoiser-callback

Add mid-kdiffusion cfgdenoiser script callback - access latents, conditionings and sigmas mid-sampling

Author: AUTOMATIC1111

modules/script_callbacks.py

@@ -26,13 +26,32 @@ def __init__(self, image, p, filename, pnginfo):
         """dictionary with parameters for image's PNG info data; infotext will have the key 'parameters'"""
 
 
+class CFGDenoiserParams:
+    def __init__(self, x, image_cond, sigma, sampling_step, total_sampling_steps):
+        self.x = x
+        """Latent image representation in the process of being denoised"""
+        
+        self.image_cond = image_cond
+        """Conditioning image"""
+        
+        self.sigma = sigma
+        """Current sigma noise step value"""
+        
+        self.sampling_step = sampling_step
+        """Current Sampling step number"""
+        
+        self.total_sampling_steps = total_sampling_steps
+        """Total number of sampling steps planned"""
+
+
 ScriptCallback = namedtuple("ScriptCallback", ["script", "callback"])
 callbacks_app_started = []
 callbacks_model_loaded = []
 callbacks_ui_tabs = []
 callbacks_ui_settings = []
 callbacks_before_image_saved = []
 callbacks_image_saved = []
+callbacks_cfg_denoiser = []
 
 
 def clear_callbacks():
@@ -41,7 +60,7 @@ def clear_callbacks():
     callbacks_ui_settings.clear()
     callbacks_before_image_saved.clear()
     callbacks_image_saved.clear()
-
+    callbacks_cfg_denoiser.clear()
 
 def app_started_callback(demo: Blocks, app: FastAPI):
     for c in callbacks_app_started:
@@ -95,6 +114,14 @@ def image_saved_callback(params: ImageSaveParams):
             report_exception(c, 'image_saved_callback')
 
 
+def cfg_denoiser_callback(params: CFGDenoiserParams):
+    for c in callbacks_cfg_denoiser:
+        try:
+            c.callback(params)
+        except Exception:
+            report_exception(c, 'cfg_denoiser_callback')
+
+
 def add_callback(callbacks, fun):
     stack = [x for x in inspect.stack() if x.filename != __file__]
     filename = stack[0].filename if len(stack) > 0 else 'unknown file'
@@ -147,3 +174,12 @@ def on_image_saved(callback):
         - params: ImageSaveParams - parameters the image was saved with. Changing fields in this object does nothing.
     """
     add_callback(callbacks_image_saved, callback)
+
+
+def on_cfg_denoiser(callback):
+    """register a function to be called in the kdiffussion cfg_denoiser method after building the inner model inputs.
+    The callback is called with one argument:
+        - params: CFGDenoiserParams - parameters to be passed to the inner model and sampling state details.
+    """
+    add_callback(callbacks_cfg_denoiser, callback)
+

modules/sd_samplers.py

@@ -12,6 +12,7 @@
 
 from modules.shared import opts, cmd_opts, state
 import modules.shared as shared
+from modules.script_callbacks import CFGDenoiserParams, cfg_denoiser_callback
 
 
 SamplerData = namedtuple('SamplerData', ['name', 'constructor', 'aliases', 'options'])
@@ -280,6 +281,12 @@ def forward(self, x, sigma, uncond, cond, cond_scale, image_cond):
         image_cond_in = torch.cat([torch.stack([image_cond[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [image_cond])
         sigma_in = torch.cat([torch.stack([sigma[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [sigma])
 
+        denoiser_params = CFGDenoiserParams(x_in, image_cond_in, sigma_in, state.sampling_step, state.sampling_steps)
+        cfg_denoiser_callback(denoiser_params)
+        x_in = denoiser_params.x
+        image_cond_in = denoiser_params.image_cond
+        sigma_in = denoiser_params.sigma
+
         if tensor.shape[1] == uncond.shape[1]:
             cond_in = torch.cat([tensor, uncond])