v0.4.5

OMEinsum v0.4.5

Diff since v0.4.4

Merged pull requests:

• change flatten (#111) (@GiggleLiu)

v0.4.4

OMEinsum v0.4.4

Diff since v0.4.3

Merged pull requests:

• fix several gpu binary rules (#110) (@GiggleLiu)

v0.4.3

OMEinsum v0.4.3

Diff since v0.4.2

Closed issues:

• doc wrong (#108)

Merged pull requests:

• fix greedy contraction order (#109) (@GiggleLiu)

v0.4.2

OMEinsum v0.4.2

Diff since v0.4.1

Merged pull requests:

• Faster greedy (#107) (@GiggleLiu)

v0.4.1

OMEinsum v0.4.1

Diff since v0.4.0

Merged pull requests:

• greedy algorithm for optimizing einsum contraction order (#105) (@GiggleLiu)
• better greedy (#106) (@GiggleLiu)

v0.4.0

OMEinsum v0.4.0

Diff since v0.3.3

Closed issues:

• CUDA incompatibility (#102)

Merged pull requests:

• Refactor 2021 03 (#104) (@GiggleLiu)

v0.3.3

OMEinsum v0.3.3

Diff since v0.3.2

Closed issues:

• Evaluate Hessian of an einsum with Zygote.hessian (#98)
• zygote + OMEinsum fails (#99)

Merged pull requests:

• upgrade to CuArray2.1 (#96) (@GiggleLiu)
• Upgrad deps (#100) (@GiggleLiu)
• Fix a debug message (#101) (@mcabbott)

v0.3.2

OMEinsum v0.3.2

Diff since v0.3.1

Closed issues:

• Type inference failures (#92)
• Make permutedims faster (#94)

Merged pull requests:

• fix a bug in #80 (#85) (@GiggleLiu)
• Install TagBot as a GitHub Action (#86) (@JuliaTagBot)
• Add debug macro logging (#87) (@mcabbott)
• fix type instability (#93) (@GiggleLiu)
• Faster permutedims using TensorOperations (#95) (@mcabbott)

v0.3.1

v0.3.1 (2020-02-07)

Diff since v0.3.0

Closed issues:

• error with high precision numbers on GPU (#79)
• Einsum fails on certain configurations for CuArrays (#78)
• List of bad interfaces to rename (#66)
• Allowing specify missing index sizes (#44)

Merged pull requests:

• Add a note about brackets (#84) (mcabbott)
• Update Project.toml (#83) (mcabbott)
• use ZygoteRules (#82) (mcabbott)
• An option to disable fallback (#81) (mcabbott)
• Fix batchedmm dispatch (#80) (GiggleLiu)
• use CuArrays, CUDAnative, and CUDAdrv through .CuArrays (#77) (jfeist)
• use batched gemm to speed up GPU and lower the JIT time (#74) (GiggleLiu)
• allow additional size information (#67) (GiggleLiu)

v0.3.0

v0.3.0 (2019-08-26)

Diff since v0.2.0

Closed issues:

• Register 0.2.0 (#58)
• CI fail for nightly build (#57)
• Using generated function without doing any calculation during type inference (#52)
• Why isn't the notatation for einsumopt the same as for einsum? (#49)
• Optimal Contraction Order (#28)
• Add simd support to increase CPU performance (#26)
• use @adjoint! instead of @adjoint for einsum! AD (#18)
• About type changes in backward pass (#16)
• List of testing cases (#6)

Merged pull requests:

• fix scalar output for sum, trace and hadamard, fix scalar input, expo… (#72) (GiggleLiu)
• Add flowchart to index (#71) (under-Peter)
• Remove old interface, einsumopt and superfluous implementations (#70) (under-Peter)
• 3 coloring example (#68) (GiggleLiu)
• allow additional size information (#67) (GiggleLiu)
• Add docs (#65) (under-Peter)
• Add matmul permutations (#64) (under-Peter)
• Allow permtations in Hadamard/Sum/Partial Trace specifications (#62) (under-Peter)
• faster fallback implementation (#60) (GiggleLiu)