requirements-firered-tts-2.txt

# Optional FireRedTTS-2 backend dependencies.
#
# Install these only if you plan to use backend=firered-tts-2 voice profiles:
#   pip install -r requirements-firered-tts-2.txt
#
# Download weights separately; the backend does not auto-download them:
#   git lfs install
#   git clone https://huggingface.co/FireRedTeam/FireRedTTS2 \
#     backends/extras/firered-tts-2/FireRedTTS2
#
# FireRedTTS-2 code and weights are Apache-2.0. Upstream is CUDA-first; set
# FIRERED_TTS_2_DEVICE=mps/cpu/cuda to override device selection.
git+https://github.com/FireRedTeam/FireRedTTS2.git@404f3f61d25bb4804859b588a6a734bf8468090c
torch>=2.12.0
torchaudio>=2.11.0
torchvision>=0.27.1
# torchtune and torchao have macOS arm64 wheels; set FIRERED_TTS_2_DEVICE=mps
# for Metal acceleration. Backend defaults to MPS → CPU via PYTORCH_ENABLE_MPS_FALLBACK.
torchtune>=0.6.0
torchao>=0.10.0
transformers>=5.12.0
einops>=0.8.0
librosa>=0.10.2
optuna>=4.0.0
accelerate>=1.0.0
# tensorboard: training-side logging only — not imported by the afterwords server at runtime.
# Do not start the tensorboard dashboard (binds 0.0.0.0:6006 unauthenticated) in production.
tensorboard>=2.20.0