mirrors/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2025-09-12 12:37:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

More flexible long clip support.

Browse Source 
Add clip g long clip support.

Text encoder refactor.

Support llama models with different vocab sizes.
This commit is contained in:
comfyanonymous
2025-04-15 10:32:21 -04:00
parent 8a438115fb
commit 3e8155f7a3
17 changed files with 95 additions and 66 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
comfy/text_encoders/flux.py
View File

@@ -9,14 +9,13 @@ import os
class T5XXLTokenizer(sd1_clip.SDTokenizer):
def __init__(self, embedding_directory=None, tokenizer_data={}):
tokenizer_path = os.path.join(os.path.dirname(os.path.realpath(__file__)), "t5_tokenizer")
super().__init__(tokenizer_path, embedding_directory=embedding_directory, pad_with_end=False, embedding_size=4096, embedding_key='t5xxl', tokenizer_class=T5TokenizerFast, has_start_token=False, pad_to_max_length=False, max_length=99999999, min_length=256)
super().__init__(tokenizer_path, embedding_directory=embedding_directory, pad_with_end=False, embedding_size=4096, embedding_key='t5xxl', tokenizer_class=T5TokenizerFast, has_start_token=False, pad_to_max_length=False, max_length=99999999, min_length=256, tokenizer_data=tokenizer_data)
class FluxTokenizer:
def __init__(self, embedding_directory=None, tokenizer_data={}):
clip_l_tokenizer_class = tokenizer_data.get("clip_l_tokenizer_class", sd1_clip.SDTokenizer)
self.clip_l = clip_l_tokenizer_class(embedding_directory=embedding_directory)
self.t5xxl = T5XXLTokenizer(embedding_directory=embedding_directory)
self.clip_l = sd1_clip.SDTokenizer(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data)
self.t5xxl = T5XXLTokenizer(embedding_directory=embedding_directory, tokenizer_data=tokenizer_data)
def tokenize_with_weights(self, text:str, return_word_ids=False, **kwargs):
out = {}
@@ -35,8 +34,7 @@ class FluxClipModel(torch.nn.Module):
def __init__(self, dtype_t5=None, device="cpu", dtype=None, model_options={}):
super().__init__()
dtype_t5 = comfy.model_management.pick_weight_dtype(dtype_t5, dtype, device)
clip_l_class = model_options.get("clip_l_class", sd1_clip.SDClipModel)
self.clip_l = clip_l_class(device=device, dtype=dtype, return_projected_pooled=False, model_options=model_options)
self.clip_l = sd1_clip.SDClipModel(device=device, dtype=dtype, return_projected_pooled=False, model_options=model_options)
self.t5xxl = comfy.text_encoders.sd3_clip.T5XXLModel(device=device, dtype=dtype_t5, model_options=model_options)
self.dtypes = set([dtype, dtype_t5])