Sparse mode performance, SparseHist input dispatch, and low-memory --noHessian mode#129

Open

bendavid wants to merge 18 commits intoWMass:mainfrom

bendavid:sparsedev4

Commits on Apr 9, 2026

add option to treat input systematic histograms as difference with respect to nominal
bendavid
committed
add test for sparse mode
bendavid
committed
Support scipy sparse array inputs in TensorWriter and add as_difference option

bendavid
and
claude
committed
Add multi-systematic dispatch in add_systematic and use wums.SparseHist

bendavid
and
claude
committed
Add external likelihood term (gradient + hessian) support

bendavid
and
claude
committed
Add efficient SparseHist multi-systematic dispatch in TensorWriter

bendavid
and
claude
committed
Speed up TensorWriter for large multi-systematic SparseHist workloads

bendavid
and
claude
committed
inputdata, parsing: prep for sparse fast path with CSR matvec

bendavid
and
claude
committed
fitter: dynamic loss/grad/HVP wrappers with jit_compile + hvpMethod

bendavid
and
claude
committed
fitter: sparse fast path uses CSR matmul, no dense [nbins,nproc]

bendavid
and
claude
committed
fitter: external sparse Hessian via CSR matmul

bendavid
and
claude
committed
rabbit_fit, setup.sh: enable XLA multi-threaded Eigen on CPU

bendavid
and
claude
committed
fitter, rabbit_fit: skip dense cov allocation under --noHessian

bendavid
and
claude
committed
fitter: speed up Fitter.__init__ on large external sparse Hessians

bendavid
and
claude
committed
unify sparse-Hessian IO path; sort at write time, drop reorder calls

bendavid
and
claude
committed
fitter: Hessian-free CG solve for is_linear case under --noHessian

bendavid
and
claude
committed

Commits on Apr 10, 2026