wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-01-16 01:00:49 +08:00
Code Issues Actions 13 Packages Projects Releases Wiki Activity

.

Browse Source
This commit is contained in:
Yousef Rafat 2025-12-05 22:54:51 +02:00
parent 95803e3537
commit 0ffc80f167
2 changed files with 9 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
comfy/text_encoders/hunyuan_image.py
View File

@ -13,8 +13,9 @@ class HunyuanImage3TextEncoder(torch.nn.Module):
self.num_layers = 1
def get_input_embeddings(self):
return self.wte
def forward(self, x):
return x, torch.empty_like(x)
def forward(self, *args, **kwargs):
embeds = kwargs.get("embeds")
return embeds, torch.empty_like(embeds)
class HunyuanImage3(sd1_clip.SDClipModel):
def __init__(self, device="cpu", max_length=77, freeze=True, layer="last", layer_idx=None, textmodel_json_config=None, dtype=None, model_class=HunyuanImage3TextEncoder, layer_norm_hidden_state=True, enable_attention_masks=False, zero_out_masked=False, return_projected_pooled=False, return_attention_masks=False, model_options={}):
@ -22,9 +23,9 @@ class HunyuanImage3(sd1_clip.SDClipModel):
super().__init__(device=device, max_length=max_length, freeze=freeze, layer=layer, layer_idx=layer_idx, textmodel_json_config=textmodel_json_config, dtype=dtype, model_class=model_class, layer_norm_hidden_state=layer_norm_hidden_state, enable_attention_masks=enable_attention_masks, zero_out_masked=zero_out_masked, return_projected_pooled=return_projected_pooled, return_attention_masks=return_attention_masks, model_options=model_options, special_tokens={"pad": 128009})
class HunyuanImage3Tokenizer(sd1_clip.SDTokenizer):
def __init__(self, tokenizer_path="hunyuan3_tokenizer", max_length=77, pad_with_end=True, embedding_directory=None, embedding_size=4096, embedding_key='clip_l', tokenizer_class=AutoTokenizer, pad_to_max_length=False, min_length=None, pad_token=128009, tokenizer_data = {}):
def __init__(self, tokenizer_path="hunyuan3_tokenizer", max_length=77, pad_with_end=True, embedding_directory=None, embedding_size=4096, embedding_key='clip_l', tokenizer_class=AutoTokenizer, pad_to_max_length=False, pad_token=128009, tokenizer_data = {}):
tokenizer_path = os.path.join(os.path.dirname(os.path.realpath(__file__)), tokenizer_path)
super().__init__(tokenizer_path, max_length, pad_with_end, embedding_directory, embedding_size, embedding_key, tokenizer_class, pad_to_max_length, min_length, pad_token)
super().__init__(tokenizer_path=tokenizer_path, max_length=max_length, pad_with_end=pad_with_end, embedding_directory=embedding_directory, embedding_size=embedding_size, embedding_key=embedding_key, tokenizer_class=tokenizer_class, pad_to_max_length=pad_to_max_length, pad_token=pad_token, has_start_token=False, has_end_token=False)
class ByT5SmallTokenizer(sd1_clip.SDTokenizer):
def __init__(self, embedding_directory=None, tokenizer_data={}):

8
comfy_extras/nodes_hunyuan_image.py
View File

@ -49,17 +49,17 @@ class HunyuanImage3Conditioning(io.ComfyNode):
)
@classmethod
def execute(cls, text_encoding, model, text_encoding_negative=None, vae_encoding = None, vit_encoding = None):
model = model.diffusion_model
def execute(cls, text_encoding_positive, model, text_encoding_negative=None, vae_encoding = None, vit_encoding = None):
model = model.model.diffusion_model
encode_fn = model.encode_tok
special_fn = model.special_tok
word_embed = model.wte
word_embed = model.model.wte
patch_embed = model.patch_embed
t_embed = model.time_embed
text_tokens = text_encoding[0][0]
text_tokens = text_encoding_positive[0][0]
batch_size, _, hidden_size = text_tokens.shape
def fn(string, func = encode_fn):