Branch | Commit message | Author | Age | |
---|---|---|---|---|
master | Fairer benchmark: compile ad with +ffi | Tom Smeding | 4 weeks | |
Age | Commit message | Author | ||
2025-03-04 | Fairer benchmark: compile ad with +ffiHEADmaster | Tom Smeding | ||
2025-02-24 | Dual arrays is >100x faster than 'ad' on large fneural | Tom Smeding | ||
2025-02-24 | WIP array stuff untested | Tom Smeding | ||
2025-02-24 | WIP array AD | Tom Smeding | ||
2025-02-23 | Remove redundant Num constraint on writeTape | Tom Smeding | ||
2025-02-23 | Optimise by backpropagating scalar computation in C | Tom Smeding | ||
2025-02-21 | Working tests and benchmarks against 'ad' | Tom Smeding | ||
2025-02-21 | Neural returns 1 on all-zero input, not 0. | Tom Smeding | ||
2025-02-21 | Evaluate result Dual before backpropagating | Tom Smeding | ||
2025-02-21 | Nicer API for writing Dual instances | Tom Smeding | ||
[...] |