The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Add context windows callback for custom cond handling (#11208)
drozbay77b2f7c228a0
diff --git a/comfy/context_windows.py b/comfy/context_windows.pyindex 5c412d1c2..2979b3ca1 100644--- a/comfy/context_windows.py+++ b/comfy/context_windows.py@@ -87,6 +87,7 @@ class IndexListCallbacks:COMBINE_CONTEXT_WINDOW_RESULTS = "combine_context_window_results"EXECUTE_START = "execute_start"EXECUTE_CLEANUP = "execute_cleanup"+ RESIZE_COND_ITEM = "resize_cond_item"def init_callbacks(self):return {}@@ -166,6 +167,18 @@ class IndexListContextHandler(ContextHandlerABC):new_cond_item = cond_item.copy()# when in dictionary, look for tensors and CONDCrossAttn [comfy/conds.py] (has cond attr that is a tensor)for cond_key, cond_value in new_cond_item.items():+ # Allow callbacks to handle custom conditioning items+ handled = False+ for callback in comfy.patcher_extension.get_all_callbacks(+ IndexListCallbacks.RESIZE_COND_ITEM, self.callbacks+ ):+ result = callback(cond_key, cond_value, window, x_in, device, new_cond_item)+ if result is not None:+ new_cond_item[cond_key] = result+ handled = True+ break+ if handled:+ continueif isinstance(cond_value, torch.Tensor):if (self.dim < cond_value.ndim and cond_value(self.dim) == x_in.size(self.dim)) or \(cond_value.ndim < self.dim and cond_value.size(0) == x_in.size(self.dim)):