The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Support zeroing out text embeddings with the attention mask.
comfyanonymous742d5720d1b1
diff --git a/comfy/sd1_clip.py b/comfy/sd1_clip.pyindex e7ebf046d..2729f14d8 100644--- a/comfy/sd1_clip.py+++ b/comfy/sd1_clip.py@@ -68,7 +68,8 @@ class SDClipModel(torch.nn.Module, ClipTokenWeightEncoder):]def __init__(self, version="openai/clip-vit-large-patch14", device="cpu", max_length=77,freeze=True, layer="last", layer_idx=None, textmodel_json_config=None, dtype=None, model_class=comfy.clip_model.CLIPTextModel,- special_tokens={"start": 49406, "end": 49407, "pad": 49407}, layer_norm_hidden_state=True, enable_attention_masks=False, return_projected_pooled=True): # clip-vit-base-patch32+ special_tokens={"start": 49406, "end": 49407, "pad": 49407}, layer_norm_hidden_state=True, enable_attention_masks=False, zero_out_masked=False,+ return_projected_pooled=True): # clip-vit-base-patch32super().__init__()assert layer in self.LAYERS@@ -90,6 +91,7 @@ class SDClipModel(torch.nn.Module, ClipTokenWeightEncoder):self.logit_scale = torch.nn.Parameter(torch.tensor(4.6055))self.enable_attention_masks = enable_attention_masks+ self.zero_out_masked = zero_out_maskedself.layer_norm_hidden_state = layer_norm_hidden_stateself.return_projected_pooled = return_projected_pooled@@ -179,9 +181,12 @@ class SDClipModel(torch.nn.Module, ClipTokenWeightEncoder):self.transformer.set_input_embeddings(backup_embeds)if self.layer == "last":- z = outputs[0]+ z = outputs[0].float()else:- z = outputs[1]+ z = outputs[1].float()++ if self.zero_out_masked and attention_mask is not None:+ z *= attention_mask.unsqueeze(-1).float()pooled_output = Noneif len(outputs) >= 3:@@ -190,7 +195,7 @@ class SDClipModel(torch.nn.Module, ClipTokenWeightEncoder):elif outputs[2] is not None:pooled_output = outputs[2].float()- return z.float(), pooled_output+ return z, pooled_outputdef encode(self, tokens):return self(tokens)