jinx/ComfyUI
1
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2025-08-02 23:14:49 +08:00
Code Issues Activity

Move text_projection to base clip model.

Browse Source
This commit is contained in:
comfyanonymous
2023-08-24 22:20:30 -04:00
parent 30eb92c3cb
commit ec96f6d03a
4 changed files with 8 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
comfy/sd1_clip.py
View File

@@ -66,7 +66,9 @@ class SD1ClipModel(torch.nn.Module, ClipTokenWeightEncoder):
self.layer = layer
self.layer_idx = None
self.empty_tokens = [[49406] + [49407] * 76]
self.text_projection = None
self.text_projection = torch.nn.Parameter(torch.eye(self.transformer.get_input_embeddings().weight.shape[1]))
self.logit_scale = torch.nn.Parameter(torch.tensor(4.6055))
self.layer_norm_hidden_state = True
if layer == "hidden":
assert layer_idx is not None
@@ -163,6 +165,10 @@ class SD1ClipModel(torch.nn.Module, ClipTokenWeightEncoder):
return self(tokens)
def load_sd(self, sd):
if "text_projection" in sd:
self.text_projection[:] = sd.pop("text_projection")
if "text_projection.weight" in sd:
self.text_projection[:] = sd.pop("text_projection.weight").transpose(0, 1)
return self.transformer.load_state_dict(sd, strict=False)
def parse_parentheses(string):