Chase Riley Roberts
Chase Riley Roberts
It would be nice if the contraction algorithms (especially `optimal`) was written in Cython. This could give really big speed boosts to the search algorithms.
This is only a draft PR, and shouldn't be seriously reviewed until after https://github.com/openxla/xla/pull/8968 is merged into openXLA. Add a new jax.lax op `pcollective_broadcast` that calls the new `CollectiveBroadcast` StableHlo...
Adds many of the types needed when interpreting a jaxpr.
Right now, if you try to pickle a jaxpr, you are given an error. ```python >>> import jax >>> j = jax.make_jaxpr(f)(1) >>> j { lambda ; a:i32[]. let b:i32[]...
Channel handle was never used previously. This change is dependent on https://github.com/openxla/xla/pull/11418
Noticed the new nccl api wasn't documented.
As discussed in https://github.com/google/TensorNetwork/issues/179, we should merge axes names after contraction and throw an error if the user tries to access an ambiguous name.
To help with user adoption, we need additional tutorials beyond what we have in the README. Some nice things to have include: * Quantum computing example * SAT solver example...
We should consider renaming the methods around NconBuilder. Current changes proposed: NconBuilder -> ContractionBuilder tn.contract -> tn.contract_edge tn.finalize -> tn.contract
We should develop a way to easily/cheaply calculate all of the environments of a tensor network without any dangling edges.