Pinned Fiats tag 0.12.0 (upstream manifest version 0.11.1), named inference-engine at that revision, with an explicitly pinned compatible Julienne 1.3.1 dependency and Assert 1.7.0. Tag 0.14.0 was tried first but its consumer build cannot find Julienne modules because they are declared only as a development dependency. Single-image coarrays and long free-form lines are enabled. Based on the author's write-read-infer example, the checked consumer constructs a two-input ReLU network with known diagonal weights and biases, evaluates negative, zero and positive input pairs, and compares every output with direct formulas. K is a broad function-approximation classification; training algorithms, gradients, model-file round trips and parallel speedups are not verified. The historical version is explicit in the menu description. The original Berkeley Lab license terms, including the enhancements clause, are retained.
Package dependencies
Resolved GitHub repositories used by the verified playground build, combining direct and indirect dependencies. This list does not describe internal module dependencies or native system libraries.
Pinned Fiats tag 0.12.0 (upstream manifest version 0.11.1), named inference-engine at that revision, with an explicitly pinned compatible Julienne 1.3.1 dependency and Assert 1.7.0. Tag 0.14.0 was tried first but its consumer build cannot find Julienne modules because they are declared only as a development dependency. Single-image coarrays and long free-form lines are enabled. Based on the author's write-read-infer example, the checked consumer constructs a two-input ReLU network with known diagonal weights and biases, evaluates negative, zero and positive input pairs, and compares every output with direct formulas. K is a broad function-approximation classification; training algorithms, gradients, model-file round trips and parallel speedups are not verified. The historical version is explicit in the menu description. The original Berkeley Lab license terms, including the enhancements clause, are retained. The dedicated GAMS driver additionally checks 48 fixed multilayer networks with irregular widths, three activations, scalar references, padding sentinels and repeated inference; default identity input/output ranges only.