wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-06-24 00:39:30 +08:00
Code Issues Actions 18 Packages Projects Releases Wiki Activity

Remove prematurely added image embedding code from the Ideogram 4 text encoder

Browse Source
This commit is contained in:
silveroxides 2026-06-15 09:35:49 +02:00
parent 213c7d8914
commit ff05eb9a78
1 changed files with 2 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
comfy/text_encoders/ideogram4.py
View File

@ -32,14 +32,11 @@ class Ideogram4Tokenizer(sd1_clip.SD1Tokenizer):
name="qwen3vl_8b", tokenizer=Qwen3VLTokenizer) name="qwen3vl_8b", tokenizer=Qwen3VLTokenizer)
self.llama_template = "<|im_start|>user\n{}<|im_end|>\n<|im_start|>assistant\n" self.llama_template = "<|im_start|>user\n{}<|im_end|>\n<|im_start|>assistant\n"
self.llama_template_images = "<|im_start|>user\n<|vision_start|><|image_pad|><|vision_end|>{}<|im_end|>\n<|im_start|>assistant\n"
def tokenize_with_weights(self, text, return_word_ids=False, llama_template=None, images=[], prevent_empty_text=False, **kwargs): def tokenize_with_weights(self, text, return_word_ids=False, llama_template=None, prevent_empty_text=False, **kwargs):
skip_template = False skip_template = False
if text.startswith('<|im_start|>'): if text.startswith('<|im_start|>'):
skip_template = True skip_template = True
if text.startswith('<|start_header_id|>'):
skip_template = True
if prevent_empty_text and text == '': if prevent_empty_text and text == '':
text = ' ' text = ' '
@ -47,23 +44,11 @@ class Ideogram4Tokenizer(sd1_clip.SD1Tokenizer):
llama_text = text llama_text = text
else: else:
if llama_template is None: if llama_template is None:
if len(images) > 0: llama_text = self.llama_template.format(text)
llama_text = self.llama_template_images.format(text)
else:
llama_text = self.llama_template.format(text)
else: else:
llama_text = llama_template.format(text) llama_text = llama_template.format(text)
tokens = super().tokenize_with_weights(llama_text, return_word_ids=return_word_ids, disable_weights=True, **kwargs) tokens = super().tokenize_with_weights(llama_text, return_word_ids=return_word_ids, disable_weights=True, **kwargs)
key_name = next(iter(tokens))
embed_count = 0
qwen_tokens = tokens[key_name]
for r in qwen_tokens:
for i in range(len(r)):
if r[i][0] == 151655:
if len(images) > embed_count:
r[i] = ({"type": "image", "data": images[embed_count], "original_type": "image"},) + r[i][1:]
embed_count += 1
return tokens return tokens