BoltzmannEntropy / xtts2-ui

A User Interface for XTTS-2 Text-Based Voice Cloning using only 10 seconds of speech
MIT License
259 stars 42 forks source link

There is support for FlashAttention2? I have successfully install and update all from all but yet I get this message below |INSIDE| #31

Open search620 opened 3 months ago

search620 commented 3 months ago

ERROR:

xtts2-ui.venv\lib\site-packages\transformers\models\gpt2\modeling_gpt2.py:650: UserWarning: 1Torch was not compiled with flash attention. (Triggered internally at ..\aten\src\ATen\native\transformers\cuda\sdp_utils.cpp:455.) attn_output = torch.nn.functional.scaled_dot_product_attention(

also by this script I get the results below:

import torch
import transformers
from TTS.api import TTS

def check_environment():
    print("Checking environment for Flash Attention 2 compatibility...")

    # Check CUDA availability
    print(f"CUDA available: {torch.cuda.is_available()}")
    if torch.cuda.is_available():
        print(f"CUDA version: {torch.version.cuda}")

    # Check PyTorch version
    print(f"PyTorch version: {torch.__version__}")

    # Check Transformers version
    print(f"Transformers version: {transformers.__version__}")

    # Check for Flash Attention 2
    if hasattr(torch.nn.functional, 'scaled_dot_product_attention'):
        print("Flash Attention 2 is available!")
    else:
        print("Flash Attention 2 is not available.")

    # Check TTS installation
    try:
        tts = TTS(model_name="tts_models/multilingual/multi-dataset/xtts_v2")
        print("TTS model loaded successfully.")
    except Exception as e:
        print(f"Error loading TTS model: {e}")

    print("Environment check complete.")

if __name__ == "__main__":
    check_environment()

Results from the script above:

Checking environment for Flash Attention 2 compatibility... CUDA available: True CUDA version: 12.1 PyTorch version: 2.3.1+cu121 Transformers version: 4.42.4 Flash Attention 2 is available! tts_models/multilingual/multi-dataset/xtts_v2 is already downloaded. Using model: xtts TTS model loaded successfully. Environment check complete.

There solution to that?