Skip to content

Rainbow Tensor 1.1.0

Choose a tag to compare

View all tags
@Niox1337 Niox1337 released this 21 Sep 18:32
· 116 commits to main since this release
v1.1.0
27685a8
This commit was signed with the committer’s verified signature.
Niox1337 Zhixiang Feng
GPG key ID: 8B4A6CFD532BC24A
Verified
Learn about vigilant mode.

Rainbow Tensor 1.1.0 makes tensor results easier to inspect and explain in notebooks.

Highlights:

  • Compare an indexed source with its ordered result, including repeated picks, reverse slices, scalar results, and empty results.
  • Trace individual outputs of sum, mean, matrix multiplication, and einsum back to their contributing coordinates.
  • Explore output coordinates with optional notebook controls and export the current view as SVG.
  • Inspect available strides, contiguity, and ownership metadata.
  • Keep large previews bounded with compact index mappings and per-output calculation limits.

Compatibility:

  • Index selections are compact iterables. Convert them to a list only when all coordinates are needed.
  • Numerical previews use Python scalar arithmetic, which can differ from backend dtype accumulation, rounding, and overflow.
  • Scalar and zero-sized source arrays remain unsupported. Scalar and empty indexing results are supported.

Install:

pip install --upgrade rainbow-tensor==1.1.0
pip install --upgrade "rainbow-tensor[interactive]==1.1.0"

Validation includes the full local regression suite, executed tutorials, wheel installation, and independent CPU checks for PyTorch, JAX, and TensorFlow.

Assets 4
Loading