code/requirements.txt

# PyTorch with CUDA 12.8 (for Windows/Linux)
--extra-index-url https://download.pytorch.org/whl/cu128
torch==2.7.1; sys_platform == 'win32'
torchaudio==2.7.1; sys_platform == 'win32'
torchvision; sys_platform == 'win32'
# macOS arm64 (Apple Silicon): latest CPU/MPS wheels
torch>=2.9.1; sys_platform == 'darwin' and platform_machine == 'arm64'
torchaudio>=2.9.1; sys_platform == 'darwin' and platform_machine == 'arm64'
torchvision; sys_platform == 'darwin' and platform_machine == 'arm64'
# Other non-Windows platforms
torch==2.7.1; sys_platform != 'win32' and sys_platform != 'darwin'
torchaudio==2.7.1; sys_platform != 'win32' and sys_platform != 'darwin'
torchvision; sys_platform != 'win32' and sys_platform != 'darwin'

# Core dependencies
transformers>=4.51.0,<4.58.0
diffusers
gradio
matplotlib>=3.7.5
scipy>=1.10.1
soundfile>=0.13.1
loguru>=0.7.3
einops>=0.8.1
accelerate>=1.12.0
fastapi>=0.110.0
uvicorn[standard]>=0.27.0
numba>=0.63.1
vector-quantize-pytorch>=1.27.15
torchcodec>=0.9.1; sys_platform != 'darwin' or platform_machine == 'arm64'

# LoRA Training dependencies (optional)
peft>=0.7.0
lightning>=2.0.0

# nano-vllm dependencies
triton-windows>=3.0.0,<3.4; sys_platform == 'win32'
triton>=3.0.0; sys_platform != 'win32'
flash-attn @ https://github.com/sdbds/flash-attention-for-windows/releases/download/2.8.2/flash_attn-2.8.2+cu128torch2.7.1cxx11abiFALSEfullbackward-cp311-cp311-win_amd64.whl ; sys_platform == 'win32' and python_version == '3.11' and platform_machine == 'AMD64'
flash-attn; sys_platform != 'win32'
xxhash

# Local package - install with: pip install -e acestep/third_parts/nano-vllm
# nano-vllm