Merge branch 'main' into swanlab-dev

diffsynth/models/wan_video_vae.py

@@ -785,6 +785,7 @@ class WanVideoVAE(nn.Module):
                 video = self.single_decode(hidden_state, device)
             video = video.squeeze(0)
             videos.append(video)
+        videos = torch.stack(videos)
         return videos
 
 

diffsynth/prompters/wan_prompter.py

@@ -2,20 +2,24 @@ from .base_prompter import BasePrompter
 from ..models.wan_video_text_encoder import WanTextEncoder
 from transformers import AutoTokenizer
 import os, torch
+import ftfy
 import html
 import string
 import regex as re
 
 
 def basic_clean(text):
+    text = ftfy.fix_text(text)
     text = html.unescape(html.unescape(text))
     return text.strip()
 
+
 def whitespace_clean(text):
     text = re.sub(r'\s+', ' ', text)
     text = text.strip()
     return text
 
+
 def canonicalize(text, keep_punctuation_exact_string=None):
     text = text.replace('_', ' ')
     if keep_punctuation_exact_string:
@@ -28,6 +32,7 @@ def canonicalize(text, keep_punctuation_exact_string=None):
     text = re.sub(r'\s+', ' ', text)
     return text.strip()
 
+
 class HuggingfaceTokenizer:
 
     def __init__(self, name, seq_len=None, clean=None, **kwargs):

diffsynth/trainers/text_to_image.py

@@ -269,7 +269,6 @@ def launch_training_task(model, args):
         batch_size=args.batch_size,
         num_workers=args.dataloader_num_workers
     )
-
     # train
     trainer = pl.Trainer(
         max_epochs=args.max_epochs,