DaCe AD automatically differentiates scientific Python and Fortran code through an SDFG intermediate representation, beating JAX JIT across NPBench with a 4.1x geometric mean speedup.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
DaCe AD: Unifying High-Performance Automatic Differentiation for Machine Learning and Scientific Computing
DaCe AD automatically differentiates scientific Python and Fortran code through an SDFG intermediate representation, beating JAX JIT across NPBench with a 4.1x geometric mean speedup.