Skip to content

ml_dtypes

ml_dtypes is a stand-alone implementation of several NumPy dtype extensions used in machine learning libraries, including:

bfloat16: an alternative to the standard float16 format float8_*: several experimental 8-bit floating point representations including: float8_e4m3b11fnuz float8_e4m3fn float8_e4m3fnuz float8_e5m2 float8_e5m2fnuz

homepage: https://github.com/jax-ml/ml_dtypes

Available installations

ml_dtypes version Supported CPU targets Supported GPU targets EESSI version Module
0.3.2 generic: aarch64, x86_64
Arm: a64fx, neoverse_n1, neoverse_v1, nvidia/grace
AMD: zen2, zen3, zen4
Intel: haswell, skylake_avx512, sapphirerapids, icelake, cascadelake
(none) 2023.06 ml_dtypes/0.3.2-gfbf-2023a

Extensions

Overview of extensions included in ml_dtypes installations

etils

etils version ml_dtypes modules that include it
1.6.0 ml_dtypes/0.3.2-gfbf-2023a

ml_dtypes

ml_dtypes version ml_dtypes modules that include it
0.3.2 ml_dtypes/0.3.2-gfbf-2023a

opt_einsum

opt_einsum version ml_dtypes modules that include it
3.3.0 ml_dtypes/0.3.2-gfbf-2023a