accelerate>=1.15.0
torch>=2.14.1
safetensors>=0.8.0
transformers>=5.20.2
threadpoolctl>=3.7.0
packaging>=26.3
device-smi>=0.5.7
protobuf>=7.36.2
requests>=2.34.2
pypcre>=0.6.4
tokenicer>=0.0.14
logbar>=0.4.15
jinja2>=3.1.6
ninja>=1.13.2
maturin>=1.15.0
datasets>=5.1.0
dill>=0.4.1
torchao>=0.18.0
defuser>=0.0.29

[:python_version < "3.11"]
numpy>=2.2.6
filelock>=4.1.0
pillow>=12.3.0
pyarrow>=25.0.1

[:python_version >= "3.11"]
filelock>=4.2.4
pillow>=13.0.0
pyarrow>=26.0.0

[:python_version >= "3.11" and python_version < "3.12"]
numpy>=2.4.6

[:python_version >= "3.12"]
numpy>=2.5.4

[bitblas]
bitblas==0.1.0.post1

[bitsandbytes]
bitsandbytes>=0.50.2

[eval]
Evalution>=0.0.17

[gguf]
gguf>=0.19.0
llama-cpp-python>=0.3.36

[hf]
optimum>=2.3.0

[marlin-cuda]
nvidia-cuda-runtime>=13.4.92
nvidia-cublas>=13.8.1.7
nvidia-cusparse>=12.8.6.72
nvidia-cusolver>=12.3.5.3

[marlin-cuda12]
nvidia-cuda-runtime-cu12==12.9.79
nvidia-cublas-cu12==12.9.2.10
nvidia-cusparse-cu12==12.5.10.65
nvidia-cusolver-cu12==11.7.5.82

[mlx]
mlx>=0.32.4

[mlx:python_version < "3.11"]
mlx_lm>=0.31.3

[mlx:python_version >= "3.11"]
mlx_lm>=0.32.2

[openai]
uvicorn>=0.54.0
fastapi>=0.161.1
pydantic>=2.14.1

[quality]
ruff==0.14.2

[sglang]

[sglang:python_version < "3.11"]
sglang[srt]>=0.5.21

[sglang:python_version >= "3.11"]
sglang[srt]>=0.5.22

[test]
pytest>=9.1.1
pytest-timeout>=2.4.0
parameterized>=0.9.0

[triton]

[triton:python_version < "3.15"]
triton>=3.8.0

[vllm]

[vllm:python_version < "3.11"]
vllm>=0.31.0

[vllm:python_version >= "3.11" and python_version < "3.15"]
vllm>=0.32.0
