Issues

• float8_e8m0fnu: float32 values in (2^-127, 1.5*2^-127) cast to 2^-126 (code 1) instead of 2^-127 (code 0)

  jax-ml/ml_dtypes#406 · Filter by author sw23 opened on Oct 4, 2026

• Test failures on RISC-V/musl [inverse trig functions, (b)complex32]

  jax-ml/ml_dtypes#404 · Filter by author orlitzky opened on Sep 17, 2026

• MXInt8 request

  jax-ml/ml_dtypes#342 · Filter by author silobadTT opened on Dec 12, 2025

• Will this library support float4_e1m2?

  jax-ml/ml_dtypes#327 · Filter by author fhk357869050 opened on Oct 25, 2025
  1

• Numpy Serialization is not Unique for FP8 dtypes

  jax-ml/ml_dtypes#324 · Filter by author jon-chuang opened on Oct 8, 2025
  2

• Implement finfo within dtypes

  enhancement
  Feature

  jax-ml/ml_dtypes#318 · Filter by author jakevdp opened on Sep 20, 2025
  8

• Plan to support OCP Microscaling INT8 element data type?

  jax-ml/ml_dtypes#307 · Filter by author wonjeon opened on Jul 31, 2025
  1

• NaN comparison fails in np.testing.assert_equal

  jax-ml/ml_dtypes#301 · Filter by author justinchuby opened on Jul 3, 2025

• Rounding behavior of float8_e8m0fnu

  jax-ml/ml_dtypes#298 · Filter by author yuanyao-nv opened on Jun 11, 2025

• [Question / possible bug] When Saturate = false, underflow gives nan but overflow gives inf; is this standard? What's the use?

  jax-ml/ml_dtypes#291 · Filter by author SwuduSusuwu opened on May 17, 2025
  3

  

• Will this library support float4_e1m2? Look forward to support？

  enhancement
  jax-ml/ml_dtypes#281 · Filter by author the-mountain-rcj opened on Apr 11, 2025
  1

• TypeError: ufunc 'isnan' not supported for the input types

  bug
  jax-ml/ml_dtypes#272 · Filter by author carlthome opened on Mar 10, 2025
  2

  

• Incorrect behavior for out-of-range casts to FP8E4M3FN

  jax-ml/ml_dtypes#260 · Filter by author yuanyao-nv opened on Jan 25, 2025
  8

  

• bfloat16 @ bfloat16 -> float32?

  jax-ml/ml_dtypes#235 · Filter by author njzjz opened on Nov 13, 2024
  2

  

• [RFC] Change float8_e5m2 dtype descr from <f1 to <V1

  jax-ml/ml_dtypes#216 · Filter by author apivovarov opened on Oct 1, 2024
  3

  

• [Q] How to properly save and load fp8 NumPy arrays?

  jax-ml/ml_dtypes#207 · Filter by author apivovarov opened on Sep 25, 2024
  3

• np.testing.assert_array_equal() not compatible with bfloat16 when the value is nan

  jax-ml/ml_dtypes#206 · Filter by author ayaka14732 opened on Sep 23, 2024
  1

• TBox

  jax-ml/ml_dtypes#150 · Filter by author vxst opened on May 18, 2024
  1

• put a description ?

  jax-ml/ml_dtypes#149 · Filter by author stonebig opened on Apr 28, 2024
  1

• Deserialization of float8_e5m2

  jax-ml/ml_dtypes#148 · Filter by author wonjeon opened on Apr 25, 2024
  2

  

• Support for non-saturating mode for fp8

  jax-ml/ml_dtypes#147 · Filter by author wonjeon opened on Apr 22, 2024
  1

• bfloat16s (at least) throw error when given a format specifiier

  jax-ml/ml_dtypes#134 · Filter by author colehaus opened on Jan 8, 2024

• Custom floats do not respect np.set_printoptions

  bug
  jax-ml/ml_dtypes#125 · Filter by author jakevdp opened on Dec 5, 2023
  

• Make a 1.0 release and follow semantic versioning for breaking changes

  enhancement
  jax-ml/ml_dtypes#111 · Filter by author hawkinsp opened on Oct 6, 2023

• Deprecate creation of dtypes via np.dtype('typename')

  jax-ml/ml_dtypes#93 · Filter by author jakevdp opened on Sep 13, 2023
  4