Update requirements.txt

requirements.txt

@@ -1,32 +1,9 @@
-
-torch==2.6.0
-torchvision==0.21.0
-torchaudio==2.6.0
-opencv-python==4.9.0.80
-#diffusers==0.31.0
-transformers==4.49.0
-peft
-xformers
-tokenizers==0.21
-accelerate==1.1.1
-tqdm
-imageio
-easydict
-ftfy
-dashscope
-imageio-ffmpeg
-xfuser==0.4.1
-pyloudnorm
-soundfile
-librosa
-einops
-numpy<2
-
 # Common
 einops                  # Tensor operations
+flash-attn>=2.5.8       # CUDA accelerated attention
 
 # Training
-#torch==2.4.0                 # Torch
+torch==2.4.0                 # Torch
 omegaconf                # Configuration system.
 lpips                   # LPIPS model for VAE training
 
@@ -40,7 +17,7 @@ sentencepiece          # For Text Encoder
 diffusers              # Diffusers
 rotary-embedding-torch   # Rotary positional embedding
 transformers         # Transformers
-#torchvision==0.19.0          # Torchvision
+torchvision==0.19.0          # Torchvision
 tiktoken               # Tiktoken for generation
 transformers_stream_generator #LLM generation support
 
@@ -68,7 +45,4 @@ isort                 # Import sorting
 pre-commit              # Pre-commit hooks
 expecttest             # Pytorch dist tests
 hypothesis           # Fix randomness
-av                   # Process audio and video
-git+https://github.com/NVIDIA/apex.git
-#https://huggingface.co/ByteDance-Seed/SeedVR2-3B/resolve/main/apex-0.1-cp310-cp310-linux_x86_64.whl
-#https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
+av                   # Process audio and video